Insert code cell below
⌘/Ctrl+M B
Add text cell
Toggle header visibility
Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz 170500096/170498071 [==============================] - 2s 0us/step
train_images: (50000, 32, 32, 3) train_labels: (50000, 1) test_images: (10000, 32, 32, 3) test_labels: (10000, 1)
array([[[[ 59, 62, 63],
[ 43, 46, 45],
[ 50, 48, 43],
...,
[158, 132, 108],
[152, 125, 102],
[148, 124, 103]],
[[ 16, 20, 20],
[ 0, 0, 0],
[ 18, 8, 0],
...,
[123, 88, 55],
[119, 83, 50],
[122, 87, 57]],
[[ 25, 24, 21],
[ 16, 7, 0],
[ 49, 27, 8],
...,
[118, 84, 50],
[120, 84, 50],
[109, 73, 42]],
...,
[[208, 170, 96],
[201, 153, 34],
[198, 161, 26],
...,
[160, 133, 70],
[ 56, 31, 7],
[ 53, 34, 20]],
[[180, 139, 96],
[173, 123, 42],
[186, 144, 30],
...,
[184, 148, 94],
[ 97, 62, 34],
[ 83, 53, 34]],
[[177, 144, 116],
[168, 129, 94],
[179, 142, 87],
...,
[216, 184, 140],
[151, 118, 84],
[123, 92, 72]]],
[[[154, 177, 187],
[126, 137, 136],
[105, 104, 95],
...,
[ 91, 95, 71],
[ 87, 90, 71],
[ 79, 81, 70]],
[[140, 160, 169],
[145, 153, 154],
[125, 125, 118],
...,
[ 96, 99, 78],
[ 77, 80, 62],
[ 71, 73, 61]],
[[140, 155, 164],
[139, 146, 149],
[115, 115, 112],
...,
[ 79, 82, 64],
[ 68, 70, 55],
[ 67, 69, 55]],
...,
[[175, 167, 166],
[156, 154, 160],
[154, 160, 170],
...,
[ 42, 34, 36],
[ 61, 53, 57],
[ 93, 83, 91]],
[[165, 154, 128],
[156, 152, 130],
[159, 161, 142],
...,
[103, 93, 96],
[123, 114, 120],
[131, 121, 131]],
[[163, 148, 120],
[158, 148, 122],
[163, 156, 133],
...,
[143, 133, 139],
[143, 134, 142],
[143, 133, 144]]],
[[[255, 255, 255],
[253, 253, 253],
[253, 253, 253],
...,
[253, 253, 253],
[253, 253, 253],
[253, 253, 253]],
[[255, 255, 255],
[255, 255, 255],
[255, 255, 255],
...,
[255, 255, 255],
[255, 255, 255],
[255, 255, 255]],
[[255, 255, 255],
[254, 254, 254],
[254, 254, 254],
...,
[254, 254, 254],
[254, 254, 254],
[254, 254, 254]],
...,
[[113, 120, 112],
[111, 118, 111],
[105, 112, 106],
...,
[ 72, 81, 80],
[ 72, 80, 79],
[ 72, 80, 79]],
[[111, 118, 110],
[104, 111, 104],
[ 99, 106, 98],
...,
[ 68, 75, 73],
[ 70, 76, 75],
[ 78, 84, 82]],
[[106, 113, 105],
[ 99, 106, 98],
[ 95, 102, 94],
...,
[ 78, 85, 83],
[ 79, 85, 83],
[ 80, 86, 84]]],
...,
[[[ 35, 178, 235],
[ 40, 176, 239],
[ 42, 176, 241],
...,
[ 99, 177, 219],
[ 79, 147, 197],
[ 89, 148, 189]],
[[ 57, 182, 234],
[ 44, 184, 250],
[ 50, 183, 240],
...,
[156, 182, 200],
[141, 177, 206],
[116, 149, 175]],
[[ 98, 197, 237],
[ 64, 189, 252],
[ 69, 192, 245],
...,
[188, 195, 206],
[119, 135, 147],
[ 61, 79, 90]],
...,
[[ 73, 79, 77],
[ 53, 63, 68],
[ 54, 68, 80],
...,
[ 17, 40, 64],
[ 21, 36, 51],
[ 33, 48, 49]],
[[ 61, 68, 75],
[ 55, 70, 86],
[ 57, 79, 103],
...,
[ 24, 48, 72],
[ 17, 35, 53],
[ 7, 23, 32]],
[[ 44, 56, 73],
[ 46, 66, 88],
[ 49, 77, 105],
...,
[ 27, 52, 77],
[ 21, 43, 66],
[ 12, 31, 50]]],
[[[189, 211, 240],
[186, 208, 236],
[185, 207, 235],
...,
[175, 195, 224],
[172, 194, 222],
[169, 194, 220]],
[[194, 210, 239],
[191, 207, 236],
[190, 206, 235],
...,
[173, 192, 220],
[171, 191, 218],
[167, 190, 216]],
[[208, 219, 244],
[205, 216, 240],
[204, 215, 239],
...,
[175, 191, 217],
[172, 190, 216],
[169, 191, 215]],
...,
[[207, 199, 181],
[203, 195, 175],
[203, 196, 173],
...,
[135, 132, 127],
[162, 158, 150],
[168, 163, 151]],
[[198, 190, 170],
[189, 181, 159],
[180, 172, 147],
...,
[178, 171, 160],
[175, 169, 156],
[175, 169, 154]],
[[198, 189, 173],
[189, 181, 162],
[178, 170, 149],
...,
[195, 184, 169],
[196, 189, 171],
[195, 190, 171]]],
[[[229, 229, 239],
[236, 237, 247],
[234, 236, 247],
...,
[217, 219, 233],
[221, 223, 234],
[222, 223, 233]],
[[222, 221, 229],
[239, 239, 249],
[233, 234, 246],
...,
[223, 223, 236],
[227, 228, 238],
[210, 211, 220]],
[[213, 206, 211],
[234, 232, 239],
[231, 233, 244],
...,
[220, 220, 232],
[220, 219, 232],
[202, 203, 215]],
...,
[[150, 143, 135],
[140, 135, 127],
[132, 127, 120],
...,
[224, 222, 218],
[230, 228, 225],
[241, 241, 238]],
[[137, 132, 126],
[130, 127, 120],
[125, 121, 115],
...,
[181, 180, 178],
[202, 201, 198],
[212, 211, 207]],
[[122, 119, 114],
[118, 116, 110],
[120, 116, 111],
...,
[179, 177, 173],
[164, 164, 162],
[163, 163, 161]]]], dtype=uint8)First ten labels training dataset: [[6] [9] [9] [4] [1] [1] [2] [7] [8] [3]] This output the numeric label, need to convert to item description
((50000, 32, 32, 3), (10000, 32, 32, 3))
((3000, 32, 32, 3), (3000, 1))
((47000, 32, 32, 3), (47000, 1))
Epoch 1/200 92/92 [==============================] - 2s 7ms/step - loss: 2.8106 - accuracy: 0.2552 - val_loss: 2.3320 - val_accuracy: 0.3117 Epoch 2/200 92/92 [==============================] - 0s 5ms/step - loss: 2.2272 - accuracy: 0.3516 - val_loss: 2.1146 - val_accuracy: 0.3650 Epoch 3/200 92/92 [==============================] - 0s 5ms/step - loss: 2.0765 - accuracy: 0.3777 - val_loss: 1.9831 - val_accuracy: 0.3970 Epoch 4/200 92/92 [==============================] - 0s 5ms/step - loss: 1.9668 - accuracy: 0.3944 - val_loss: 1.8991 - val_accuracy: 0.4060 Epoch 5/200 92/92 [==============================] - 0s 4ms/step - loss: 1.8834 - accuracy: 0.4053 - val_loss: 1.8323 - val_accuracy: 0.4150 Epoch 6/200 92/92 [==============================] - 0s 5ms/step - loss: 1.8249 - accuracy: 0.4161 - val_loss: 1.8191 - val_accuracy: 0.3987 Epoch 7/200 92/92 [==============================] - 0s 5ms/step - loss: 1.7706 - accuracy: 0.4282 - val_loss: 1.7468 - val_accuracy: 0.4300 Epoch 8/200 92/92 [==============================] - 0s 5ms/step - loss: 1.7415 - accuracy: 0.4303 - val_loss: 1.7446 - val_accuracy: 0.4243 Epoch 9/200 92/92 [==============================] - 0s 4ms/step - loss: 1.7072 - accuracy: 0.4389 - val_loss: 1.6909 - val_accuracy: 0.4417 Epoch 10/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6696 - accuracy: 0.4492 - val_loss: 1.6824 - val_accuracy: 0.4357 Epoch 11/200 92/92 [==============================] - 0s 4ms/step - loss: 1.6542 - accuracy: 0.4531 - val_loss: 1.6538 - val_accuracy: 0.4540 Epoch 12/200 92/92 [==============================] - 0s 4ms/step - loss: 1.6498 - accuracy: 0.4480 - val_loss: 1.6476 - val_accuracy: 0.4430 Epoch 13/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6307 - accuracy: 0.4546 - val_loss: 1.6700 - val_accuracy: 0.4317 Epoch 14/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6091 - accuracy: 0.4634 - val_loss: 1.6021 - val_accuracy: 0.4673 Epoch 15/200 92/92 [==============================] - 0s 4ms/step - loss: 1.6002 - accuracy: 0.4667 - val_loss: 1.6153 - val_accuracy: 0.4557 Epoch 16/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5905 - accuracy: 0.4677 - val_loss: 1.6462 - val_accuracy: 0.4430 Epoch 17/200 92/92 [==============================] - 0s 4ms/step - loss: 1.5865 - accuracy: 0.4692 - val_loss: 1.6209 - val_accuracy: 0.4580
Model: "sequential" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= flatten (Flatten) (None, 3072) 0 _________________________________________________________________ dense (Dense) (None, 384) 1180032 _________________________________________________________________ dense_1 (Dense) (None, 10) 3850 ================================================================= Total params: 1,183,882 Trainable params: 1,183,882 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 2ms/step - loss: 1.6309 - accuracy: 0.4515 test set accuracy: 45.14999985694885
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[479, 35, 52, 19, 31, 7, 39, 68, 132, 138],
[ 38, 506, 9, 18, 13, 18, 27, 53, 36, 282],
[ 78, 30, 216, 83, 151, 46, 216, 116, 19, 45],
[ 38, 23, 57, 258, 46, 139, 215, 85, 24, 115],
[ 55, 13, 74, 38, 336, 33, 253, 143, 17, 38],
[ 27, 18, 52, 159, 73, 286, 176, 128, 26, 55],
[ 6, 17, 25, 49, 85, 25, 690, 46, 11, 46],
[ 29, 27, 34, 44, 58, 44, 64, 586, 9, 105],
[105, 68, 14, 20, 19, 23, 23, 29, 481, 218],
[ 35, 122, 8, 25, 7, 15, 37, 54, 20, 677]], dtype=int32)>3
['flatten', 'dense', 'dense_1']
Epoch 1/200 92/92 [==============================] - 1s 6ms/step - loss: 2.2266 - accuracy: 0.2568 - val_loss: 1.8879 - val_accuracy: 0.3313 Epoch 2/200 92/92 [==============================] - 0s 4ms/step - loss: 1.8584 - accuracy: 0.3500 - val_loss: 1.7947 - val_accuracy: 0.3713 Epoch 3/200 92/92 [==============================] - 0s 4ms/step - loss: 1.7846 - accuracy: 0.3779 - val_loss: 1.7213 - val_accuracy: 0.3963 Epoch 4/200 92/92 [==============================] - 0s 4ms/step - loss: 1.7307 - accuracy: 0.3992 - val_loss: 1.6976 - val_accuracy: 0.4170 Epoch 5/200 92/92 [==============================] - 0s 4ms/step - loss: 1.6958 - accuracy: 0.4093 - val_loss: 1.6905 - val_accuracy: 0.3913 Epoch 6/200 92/92 [==============================] - 0s 4ms/step - loss: 1.6602 - accuracy: 0.4199 - val_loss: 1.6162 - val_accuracy: 0.4373 Epoch 7/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6351 - accuracy: 0.4254 - val_loss: 1.6302 - val_accuracy: 0.4157 Epoch 8/200 92/92 [==============================] - 0s 4ms/step - loss: 1.6026 - accuracy: 0.4398 - val_loss: 1.5811 - val_accuracy: 0.4513 Epoch 9/200 92/92 [==============================] - 0s 4ms/step - loss: 1.5791 - accuracy: 0.4500 - val_loss: 1.5780 - val_accuracy: 0.4387 Epoch 10/200 92/92 [==============================] - 0s 4ms/step - loss: 1.5577 - accuracy: 0.4566 - val_loss: 1.6056 - val_accuracy: 0.4363 Epoch 11/200 92/92 [==============================] - 0s 4ms/step - loss: 1.5434 - accuracy: 0.4586 - val_loss: 1.5560 - val_accuracy: 0.4540 Epoch 12/200 92/92 [==============================] - 0s 4ms/step - loss: 1.5231 - accuracy: 0.4663 - val_loss: 1.5263 - val_accuracy: 0.4647 Epoch 13/200 92/92 [==============================] - 0s 4ms/step - loss: 1.5087 - accuracy: 0.4730 - val_loss: 1.5418 - val_accuracy: 0.4543 Epoch 14/200 92/92 [==============================] - 0s 4ms/step - loss: 1.4913 - accuracy: 0.4812 - val_loss: 1.5406 - val_accuracy: 0.4660 Epoch 15/200 92/92 [==============================] - 0s 4ms/step - loss: 1.4863 - accuracy: 0.4792 - val_loss: 1.5060 - val_accuracy: 0.4690 Epoch 16/200 92/92 [==============================] - 0s 4ms/step - loss: 1.4726 - accuracy: 0.4835 - val_loss: 1.5005 - val_accuracy: 0.4740 Epoch 17/200 92/92 [==============================] - 0s 4ms/step - loss: 1.4531 - accuracy: 0.4921 - val_loss: 1.5131 - val_accuracy: 0.4743 Epoch 18/200 92/92 [==============================] - 0s 4ms/step - loss: 1.4454 - accuracy: 0.4943 - val_loss: 1.4860 - val_accuracy: 0.4793 Epoch 19/200 92/92 [==============================] - 0s 4ms/step - loss: 1.4286 - accuracy: 0.4994 - val_loss: 1.5115 - val_accuracy: 0.4717 Epoch 20/200 92/92 [==============================] - 0s 4ms/step - loss: 1.4149 - accuracy: 0.5053 - val_loss: 1.4775 - val_accuracy: 0.4780 Epoch 21/200 92/92 [==============================] - 0s 4ms/step - loss: 1.4155 - accuracy: 0.5020 - val_loss: 1.4615 - val_accuracy: 0.4873 Epoch 22/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4037 - accuracy: 0.5064 - val_loss: 1.4438 - val_accuracy: 0.4997 Epoch 23/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3939 - accuracy: 0.5106 - val_loss: 1.4597 - val_accuracy: 0.4883 Epoch 24/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3909 - accuracy: 0.5119 - val_loss: 1.4546 - val_accuracy: 0.4923 Epoch 25/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3772 - accuracy: 0.5159 - val_loss: 1.4418 - val_accuracy: 0.5047 Epoch 26/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3647 - accuracy: 0.5241 - val_loss: 1.4262 - val_accuracy: 0.5117 Epoch 27/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3573 - accuracy: 0.5222 - val_loss: 1.4196 - val_accuracy: 0.5087 Epoch 28/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3403 - accuracy: 0.5314 - val_loss: 1.4221 - val_accuracy: 0.5110 Epoch 29/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3516 - accuracy: 0.5256 - val_loss: 1.4211 - val_accuracy: 0.5057
Model: "sequential_1" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= flatten_1 (Flatten) (None, 3072) 0 _________________________________________________________________ dense_2 (Dense) (None, 384) 1180032 _________________________________________________________________ dense_3 (Dense) (None, 10) 3850 ================================================================= Total params: 1,183,882 Trainable params: 1,183,882 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 2ms/step - loss: 1.4496 - accuracy: 0.4892 test set accuracy: 48.91999959945679
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[440, 31, 106, 34, 53, 20, 19, 82, 156, 59],
[ 24, 566, 16, 25, 20, 12, 17, 44, 77, 199],
[ 52, 24, 366, 82, 146, 75, 95, 121, 19, 20],
[ 18, 15, 101, 344, 72, 177, 99, 95, 18, 61],
[ 33, 10, 155, 59, 427, 38, 98, 138, 22, 20],
[ 8, 11, 97, 217, 90, 330, 69, 122, 27, 29],
[ 4, 13, 90, 90, 146, 43, 520, 51, 18, 25],
[ 17, 12, 51, 57, 79, 60, 14, 651, 13, 46],
[ 62, 65, 20, 28, 38, 21, 6, 42, 656, 62],
[ 25, 149, 15, 42, 14, 27, 14, 63, 59, 592]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 1s 6ms/step - loss: 2.0320 - accuracy: 0.2773 - val_loss: 1.8437 - val_accuracy: 0.3337 Epoch 2/200 92/92 [==============================] - 0s 5ms/step - loss: 1.8163 - accuracy: 0.3616 - val_loss: 1.7548 - val_accuracy: 0.3783 Epoch 3/200 92/92 [==============================] - 0s 4ms/step - loss: 1.7165 - accuracy: 0.3983 - val_loss: 1.6428 - val_accuracy: 0.4160 Epoch 4/200 92/92 [==============================] - 0s 4ms/step - loss: 1.6398 - accuracy: 0.4223 - val_loss: 1.6231 - val_accuracy: 0.4253 Epoch 5/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5989 - accuracy: 0.4379 - val_loss: 1.5628 - val_accuracy: 0.4440 Epoch 6/200 92/92 [==============================] - 0s 4ms/step - loss: 1.5569 - accuracy: 0.4509 - val_loss: 1.5561 - val_accuracy: 0.4440 Epoch 7/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5200 - accuracy: 0.4636 - val_loss: 1.5326 - val_accuracy: 0.4590 Epoch 8/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4935 - accuracy: 0.4721 - val_loss: 1.4927 - val_accuracy: 0.4737 Epoch 9/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4582 - accuracy: 0.4829 - val_loss: 1.4689 - val_accuracy: 0.4743 Epoch 10/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4416 - accuracy: 0.4918 - val_loss: 1.4801 - val_accuracy: 0.4840 Epoch 11/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4138 - accuracy: 0.5012 - val_loss: 1.4305 - val_accuracy: 0.4907 Epoch 12/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3992 - accuracy: 0.5050 - val_loss: 1.4198 - val_accuracy: 0.4903 Epoch 13/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3763 - accuracy: 0.5130 - val_loss: 1.4187 - val_accuracy: 0.4980 Epoch 14/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3529 - accuracy: 0.5230 - val_loss: 1.3895 - val_accuracy: 0.5100 Epoch 15/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3449 - accuracy: 0.5254 - val_loss: 1.3958 - val_accuracy: 0.5110 Epoch 16/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3185 - accuracy: 0.5322 - val_loss: 1.3802 - val_accuracy: 0.5177 Epoch 17/200 92/92 [==============================] - 0s 4ms/step - loss: 1.3089 - accuracy: 0.5379 - val_loss: 1.3791 - val_accuracy: 0.5143 Epoch 18/200 92/92 [==============================] - 0s 5ms/step - loss: 1.2883 - accuracy: 0.5460 - val_loss: 1.3839 - val_accuracy: 0.5150 Epoch 19/200 92/92 [==============================] - 0s 5ms/step - loss: 1.2673 - accuracy: 0.5557 - val_loss: 1.3769 - val_accuracy: 0.5167
Model: "sequential_2" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= flatten_2 (Flatten) (None, 3072) 0 _________________________________________________________________ dense_4 (Dense) (None, 384) 1180032 _________________________________________________________________ dense_5 (Dense) (None, 192) 73920 _________________________________________________________________ dense_6 (Dense) (None, 10) 1930 ================================================================= Total params: 1,255,882 Trainable params: 1,255,882 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 2ms/step - loss: 1.3997 - accuracy: 0.5015 test set accuracy: 50.15000104904175
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[731, 63, 25, 11, 25, 13, 24, 18, 53, 37],
[ 70, 695, 19, 13, 6, 14, 16, 22, 31, 114],
[153, 29, 344, 84, 142, 59, 102, 64, 12, 11],
[ 75, 36, 84, 292, 59, 195, 136, 55, 26, 42],
[108, 21, 128, 52, 431, 42, 108, 85, 12, 13],
[ 63, 25, 83, 179, 69, 390, 75, 75, 22, 19],
[ 30, 17, 59, 69, 130, 48, 593, 27, 15, 12],
[ 92, 26, 48, 48, 88, 63, 20, 556, 11, 48],
[266, 112, 8, 19, 17, 19, 6, 18, 485, 50],
[ 77, 263, 7, 32, 13, 20, 23, 40, 27, 498]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 1s 6ms/step - loss: 1.9800 - accuracy: 0.2914 - val_loss: 1.7839 - val_accuracy: 0.3553 Epoch 2/200 92/92 [==============================] - 0s 5ms/step - loss: 1.7702 - accuracy: 0.3701 - val_loss: 1.7052 - val_accuracy: 0.3923 Epoch 3/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6758 - accuracy: 0.4021 - val_loss: 1.6422 - val_accuracy: 0.4177 Epoch 4/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6115 - accuracy: 0.4282 - val_loss: 1.5511 - val_accuracy: 0.4590 Epoch 5/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5594 - accuracy: 0.4465 - val_loss: 1.5503 - val_accuracy: 0.4553 Epoch 6/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5268 - accuracy: 0.4560 - val_loss: 1.5233 - val_accuracy: 0.4527 Epoch 7/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4939 - accuracy: 0.4693 - val_loss: 1.4990 - val_accuracy: 0.4643 Epoch 8/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4687 - accuracy: 0.4795 - val_loss: 1.4700 - val_accuracy: 0.4790 Epoch 9/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4341 - accuracy: 0.4919 - val_loss: 1.4477 - val_accuracy: 0.4773 Epoch 10/200 92/92 [==============================] - 0s 5ms/step - loss: 1.4243 - accuracy: 0.4940 - val_loss: 1.4520 - val_accuracy: 0.4717 Epoch 11/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3948 - accuracy: 0.5058 - val_loss: 1.4182 - val_accuracy: 0.4977 Epoch 12/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3766 - accuracy: 0.5096 - val_loss: 1.4154 - val_accuracy: 0.4930 Epoch 13/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3515 - accuracy: 0.5205 - val_loss: 1.4243 - val_accuracy: 0.4960 Epoch 14/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3429 - accuracy: 0.5240 - val_loss: 1.4110 - val_accuracy: 0.4987 Epoch 15/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3307 - accuracy: 0.5278 - val_loss: 1.3873 - val_accuracy: 0.5083 Epoch 16/200 92/92 [==============================] - 0s 5ms/step - loss: 1.3003 - accuracy: 0.5393 - val_loss: 1.3874 - val_accuracy: 0.5077 Epoch 17/200 92/92 [==============================] - 0s 5ms/step - loss: 1.2807 - accuracy: 0.5454 - val_loss: 1.4546 - val_accuracy: 0.4857 Epoch 18/200 92/92 [==============================] - 0s 5ms/step - loss: 1.2717 - accuracy: 0.5493 - val_loss: 1.4245 - val_accuracy: 0.4997
Model: "sequential_3" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= flatten_3 (Flatten) (None, 3072) 0 _________________________________________________________________ dense_7 (Dense) (None, 384) 1180032 _________________________________________________________________ dense_8 (Dense) (None, 192) 73920 _________________________________________________________________ dense_9 (Dense) (None, 96) 18528 _________________________________________________________________ dense_10 (Dense) (None, 10) 970 ================================================================= Total params: 1,273,450 Trainable params: 1,273,450 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 2ms/step - loss: 1.4402 - accuracy: 0.4893 test set accuracy: 48.9300012588501
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[619, 24, 72, 23, 11, 76, 10, 58, 53, 54],
[ 48, 555, 29, 29, 5, 39, 12, 43, 33, 207],
[ 73, 15, 437, 75, 51, 170, 66, 90, 8, 15],
[ 21, 7, 104, 302, 28, 365, 62, 72, 11, 28],
[ 54, 10, 243, 66, 277, 123, 85, 122, 7, 13],
[ 22, 3, 91, 144, 25, 572, 40, 80, 11, 12],
[ 8, 8, 146, 100, 58, 153, 475, 34, 4, 14],
[ 32, 9, 54, 58, 46, 131, 16, 619, 3, 32],
[222, 53, 22, 29, 10, 79, 5, 28, 446, 106],
[ 52, 122, 17, 42, 8, 57, 15, 73, 23, 591]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 18s 37ms/step - loss: 1.7844 - accuracy: 0.3575 - val_loss: 1.3712 - val_accuracy: 0.5273 Epoch 2/200 92/92 [==============================] - 3s 29ms/step - loss: 1.3449 - accuracy: 0.5219 - val_loss: 1.1917 - val_accuracy: 0.5877 Epoch 3/200 92/92 [==============================] - 3s 29ms/step - loss: 1.1762 - accuracy: 0.5882 - val_loss: 1.0742 - val_accuracy: 0.6347 Epoch 4/200 92/92 [==============================] - 3s 29ms/step - loss: 1.0663 - accuracy: 0.6269 - val_loss: 0.9765 - val_accuracy: 0.6733 Epoch 5/200 92/92 [==============================] - 3s 29ms/step - loss: 0.9972 - accuracy: 0.6525 - val_loss: 0.9596 - val_accuracy: 0.6637 Epoch 6/200 92/92 [==============================] - 3s 30ms/step - loss: 0.9350 - accuracy: 0.6755 - val_loss: 0.8773 - val_accuracy: 0.7033 Epoch 7/200 92/92 [==============================] - 3s 29ms/step - loss: 0.8780 - accuracy: 0.6920 - val_loss: 0.8470 - val_accuracy: 0.7160 Epoch 8/200 92/92 [==============================] - 3s 29ms/step - loss: 0.8235 - accuracy: 0.7155 - val_loss: 0.7987 - val_accuracy: 0.7307 Epoch 9/200 92/92 [==============================] - 3s 29ms/step - loss: 0.7815 - accuracy: 0.7279 - val_loss: 0.7893 - val_accuracy: 0.7323 Epoch 10/200 92/92 [==============================] - 3s 29ms/step - loss: 0.7501 - accuracy: 0.7396 - val_loss: 0.7685 - val_accuracy: 0.7357 Epoch 11/200 92/92 [==============================] - 3s 29ms/step - loss: 0.7067 - accuracy: 0.7534 - val_loss: 0.7794 - val_accuracy: 0.7360 Epoch 12/200 92/92 [==============================] - 3s 29ms/step - loss: 0.6716 - accuracy: 0.7659 - val_loss: 0.7211 - val_accuracy: 0.7543 Epoch 13/200 92/92 [==============================] - 3s 30ms/step - loss: 0.6369 - accuracy: 0.7782 - val_loss: 0.7137 - val_accuracy: 0.7547 Epoch 14/200 92/92 [==============================] - 3s 29ms/step - loss: 0.5966 - accuracy: 0.7921 - val_loss: 0.7208 - val_accuracy: 0.7603 Epoch 15/200 92/92 [==============================] - 3s 29ms/step - loss: 0.5599 - accuracy: 0.8066 - val_loss: 0.6875 - val_accuracy: 0.7640 Epoch 16/200 92/92 [==============================] - 3s 29ms/step - loss: 0.5255 - accuracy: 0.8169 - val_loss: 0.7267 - val_accuracy: 0.7573 Epoch 17/200 92/92 [==============================] - 3s 30ms/step - loss: 0.4965 - accuracy: 0.8280 - val_loss: 0.6871 - val_accuracy: 0.7680 Epoch 18/200 92/92 [==============================] - 3s 30ms/step - loss: 0.4576 - accuracy: 0.8432 - val_loss: 0.6955 - val_accuracy: 0.7697 Epoch 19/200 92/92 [==============================] - 3s 29ms/step - loss: 0.4275 - accuracy: 0.8505 - val_loss: 0.6957 - val_accuracy: 0.7577 Epoch 20/200 92/92 [==============================] - 3s 30ms/step - loss: 0.3950 - accuracy: 0.8617 - val_loss: 0.7068 - val_accuracy: 0.7677 Epoch 21/200 92/92 [==============================] - 3s 30ms/step - loss: 0.3676 - accuracy: 0.8722 - val_loss: 0.7026 - val_accuracy: 0.7673
Model: "sequential_4" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d (Conv2D) (None, 30, 30, 128) 3584 _________________________________________________________________ max_pooling2d (MaxPooling2D) (None, 15, 15, 128) 0 _________________________________________________________________ dropout (Dropout) (None, 15, 15, 128) 0 _________________________________________________________________ conv2d_1 (Conv2D) (None, 13, 13, 256) 295168 _________________________________________________________________ max_pooling2d_1 (MaxPooling2 (None, 6, 6, 256) 0 _________________________________________________________________ dropout_1 (Dropout) (None, 6, 6, 256) 0 _________________________________________________________________ flatten_4 (Flatten) (None, 9216) 0 _________________________________________________________________ dense_11 (Dense) (None, 384) 3539328 _________________________________________________________________ dense_12 (Dense) (None, 10) 3850 ================================================================= Total params: 3,841,930 Trainable params: 3,841,930 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 3ms/step - loss: 0.7501 - accuracy: 0.7515 test set accuracy: 75.15000104904175
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[836, 17, 32, 8, 19, 7, 9, 10, 32, 30],
[ 19, 833, 7, 3, 3, 5, 15, 2, 18, 95],
[ 69, 2, 620, 44, 76, 55, 76, 35, 10, 13],
[ 29, 4, 67, 469, 69, 201, 93, 42, 10, 16],
[ 12, 2, 52, 36, 732, 35, 56, 63, 7, 5],
[ 16, 0, 36, 134, 45, 670, 35, 50, 3, 11],
[ 4, 4, 32, 28, 28, 17, 876, 4, 4, 3],
[ 18, 3, 20, 22, 46, 52, 9, 804, 2, 24],
[ 63, 30, 18, 8, 5, 9, 10, 4, 821, 32],
[ 31, 47, 7, 6, 5, 7, 9, 14, 20, 854]], dtype=int32)>8
['conv2d', 'max_pooling2d', 'dropout', 'conv2d_1', 'max_pooling2d_1', 'dropout_1', 'flatten_4', 'dense_11', 'dense_12']
Epoch 1/200 92/92 [==============================] - 4s 37ms/step - loss: 1.9011 - accuracy: 0.2919 - val_loss: 1.5971 - val_accuracy: 0.4157 Epoch 2/200 92/92 [==============================] - 3s 36ms/step - loss: 1.4398 - accuracy: 0.4783 - val_loss: 1.2686 - val_accuracy: 0.5427 Epoch 3/200 92/92 [==============================] - 3s 36ms/step - loss: 1.2529 - accuracy: 0.5549 - val_loss: 1.0766 - val_accuracy: 0.6270 Epoch 4/200 92/92 [==============================] - 3s 36ms/step - loss: 1.1115 - accuracy: 0.6080 - val_loss: 0.9797 - val_accuracy: 0.6617 Epoch 5/200 92/92 [==============================] - 3s 36ms/step - loss: 1.0099 - accuracy: 0.6475 - val_loss: 0.8825 - val_accuracy: 0.6910 Epoch 6/200 92/92 [==============================] - 3s 36ms/step - loss: 0.9372 - accuracy: 0.6707 - val_loss: 0.8058 - val_accuracy: 0.7217 Epoch 7/200 92/92 [==============================] - 3s 36ms/step - loss: 0.8683 - accuracy: 0.6977 - val_loss: 0.7991 - val_accuracy: 0.7230 Epoch 8/200 92/92 [==============================] - 3s 36ms/step - loss: 0.8129 - accuracy: 0.7136 - val_loss: 0.7217 - val_accuracy: 0.7477 Epoch 9/200 92/92 [==============================] - 3s 36ms/step - loss: 0.7694 - accuracy: 0.7311 - val_loss: 0.7155 - val_accuracy: 0.7610 Epoch 10/200 92/92 [==============================] - 3s 35ms/step - loss: 0.7298 - accuracy: 0.7441 - val_loss: 0.6809 - val_accuracy: 0.7703 Epoch 11/200 92/92 [==============================] - 3s 36ms/step - loss: 0.6851 - accuracy: 0.7593 - val_loss: 0.6510 - val_accuracy: 0.7787 Epoch 12/200 92/92 [==============================] - 3s 35ms/step - loss: 0.6507 - accuracy: 0.7719 - val_loss: 0.6268 - val_accuracy: 0.7847 Epoch 13/200 92/92 [==============================] - 3s 36ms/step - loss: 0.6160 - accuracy: 0.7848 - val_loss: 0.6192 - val_accuracy: 0.7897 Epoch 14/200 92/92 [==============================] - 3s 36ms/step - loss: 0.5967 - accuracy: 0.7914 - val_loss: 0.6098 - val_accuracy: 0.7880 Epoch 15/200 92/92 [==============================] - 3s 36ms/step - loss: 0.5640 - accuracy: 0.8024 - val_loss: 0.6300 - val_accuracy: 0.7803 Epoch 16/200 92/92 [==============================] - 3s 35ms/step - loss: 0.5421 - accuracy: 0.8104 - val_loss: 0.6057 - val_accuracy: 0.8017 Epoch 17/200 92/92 [==============================] - 3s 35ms/step - loss: 0.5196 - accuracy: 0.8169 - val_loss: 0.5776 - val_accuracy: 0.8003 Epoch 18/200 92/92 [==============================] - 3s 35ms/step - loss: 0.4957 - accuracy: 0.8269 - val_loss: 0.5683 - val_accuracy: 0.8070 Epoch 19/200 92/92 [==============================] - 3s 36ms/step - loss: 0.4729 - accuracy: 0.8336 - val_loss: 0.5641 - val_accuracy: 0.8070 Epoch 20/200 92/92 [==============================] - 3s 36ms/step - loss: 0.4509 - accuracy: 0.8401 - val_loss: 0.5778 - val_accuracy: 0.8010 Epoch 21/200 92/92 [==============================] - 3s 36ms/step - loss: 0.4255 - accuracy: 0.8504 - val_loss: 0.5631 - val_accuracy: 0.8027
Model: "sequential_19" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_40 (Conv2D) (None, 30, 30, 128) 3584 _________________________________________________________________ max_pooling2d_34 (MaxPooling (None, 15, 15, 128) 0 _________________________________________________________________ dropout_36 (Dropout) (None, 15, 15, 128) 0 _________________________________________________________________ conv2d_41 (Conv2D) (None, 13, 13, 256) 295168 _________________________________________________________________ max_pooling2d_35 (MaxPooling (None, 6, 6, 256) 0 _________________________________________________________________ dropout_37 (Dropout) (None, 6, 6, 256) 0 _________________________________________________________________ conv2d_42 (Conv2D) (None, 4, 4, 512) 1180160 _________________________________________________________________ max_pooling2d_36 (MaxPooling (None, 2, 2, 512) 0 _________________________________________________________________ dropout_38 (Dropout) (None, 2, 2, 512) 0 _________________________________________________________________ flatten_19 (Flatten) (None, 2048) 0 _________________________________________________________________ dense_46 (Dense) (None, 384) 786816 _________________________________________________________________ dense_47 (Dense) (None, 10) 3850 ================================================================= Total params: 2,269,578 Trainable params: 2,269,578 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 3ms/step - loss: 0.6289 - accuracy: 0.7872 test set accuracy: 78.71999740600586
WARNING:tensorflow:5 out of the last 1256 calls to <function Model.make_predict_function.<locals>.predict_function at 0x7fd9b262d3b0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has experimental_relax_shapes=True option that relaxes argument shapes that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for more details. shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[775, 18, 41, 8, 30, 4, 5, 9, 85, 25],
[ 13, 881, 5, 1, 1, 4, 6, 2, 28, 59],
[ 38, 5, 715, 25, 59, 53, 51, 27, 19, 8],
[ 11, 5, 80, 531, 74, 157, 62, 36, 18, 26],
[ 9, 1, 53, 28, 756, 30, 49, 62, 10, 2],
[ 10, 3, 38, 98, 43, 713, 20, 55, 14, 6],
[ 4, 6, 33, 28, 23, 15, 872, 6, 8, 5],
[ 12, 3, 16, 17, 31, 49, 3, 848, 6, 15],
[ 22, 16, 5, 7, 5, 2, 5, 1, 916, 21],
[ 14, 55, 7, 7, 4, 2, 5, 7, 34, 865]], dtype=int32)>8
['conv2d_40', 'max_pooling2d_34', 'dropout_36', 'conv2d_41', 'max_pooling2d_35', 'dropout_37', 'conv2d_42', 'max_pooling2d_36', 'dropout_38', 'flatten_19', 'dense_46', 'dense_47']
Epoch 1/200 92/92 [==============================] - 1s 7ms/step - loss: 2.6938 - accuracy: 0.2672 - val_loss: 2.3271 - val_accuracy: 0.3400 Epoch 2/200 92/92 [==============================] - 0s 5ms/step - loss: 2.1956 - accuracy: 0.3636 - val_loss: 2.0490 - val_accuracy: 0.3920 Epoch 3/200 92/92 [==============================] - 0s 5ms/step - loss: 2.0086 - accuracy: 0.3976 - val_loss: 1.9298 - val_accuracy: 0.4087 Epoch 4/200 92/92 [==============================] - 0s 5ms/step - loss: 1.9024 - accuracy: 0.4159 - val_loss: 1.8673 - val_accuracy: 0.4087 Epoch 5/200 92/92 [==============================] - 0s 5ms/step - loss: 1.8256 - accuracy: 0.4290 - val_loss: 1.8036 - val_accuracy: 0.4280 Epoch 6/200 92/92 [==============================] - 0s 5ms/step - loss: 1.7644 - accuracy: 0.4421 - val_loss: 1.7574 - val_accuracy: 0.4300 Epoch 7/200 92/92 [==============================] - 0s 5ms/step - loss: 1.7093 - accuracy: 0.4546 - val_loss: 1.6947 - val_accuracy: 0.4480 Epoch 8/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6638 - accuracy: 0.4645 - val_loss: 1.6661 - val_accuracy: 0.4583 Epoch 9/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6421 - accuracy: 0.4707 - val_loss: 1.6521 - val_accuracy: 0.4617 Epoch 10/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6201 - accuracy: 0.4709 - val_loss: 1.6200 - val_accuracy: 0.4733 Epoch 11/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6103 - accuracy: 0.4777 - val_loss: 1.6378 - val_accuracy: 0.4537 Epoch 12/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5883 - accuracy: 0.4806 - val_loss: 1.5932 - val_accuracy: 0.4793 Epoch 13/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5524 - accuracy: 0.4956 - val_loss: 1.5712 - val_accuracy: 0.4773 Epoch 14/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5542 - accuracy: 0.4917 - val_loss: 1.5576 - val_accuracy: 0.4923 Epoch 15/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5349 - accuracy: 0.4988 - val_loss: 1.5700 - val_accuracy: 0.4820 Epoch 16/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5169 - accuracy: 0.5056 - val_loss: 1.5511 - val_accuracy: 0.4923 Epoch 17/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5026 - accuracy: 0.5119 - val_loss: 1.5799 - val_accuracy: 0.4857
Model: "sequential_6" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= flatten_6 (Flatten) (None, 3072) 0 _________________________________________________________________ dense_15 (Dense) (None, 384) 1180032 _________________________________________________________________ dense_16 (Dense) (None, 192) 73920 _________________________________________________________________ dense_17 (Dense) (None, 10) 1930 ================================================================= Total params: 1,255,882 Trainable params: 1,255,882 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 2ms/step - loss: 1.5986 - accuracy: 0.4762 test set accuracy: 47.620001435279846
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[707, 57, 24, 7, 32, 11, 8, 8, 85, 61],
[ 57, 698, 10, 7, 15, 8, 4, 13, 48, 140],
[182, 44, 273, 46, 229, 68, 48, 49, 30, 31],
[105, 70, 79, 237, 132, 160, 52, 42, 40, 83],
[114, 42, 103, 31, 552, 25, 38, 51, 27, 17],
[ 87, 42, 90, 148, 119, 343, 30, 57, 35, 49],
[ 39, 49, 78, 71, 251, 72, 361, 22, 19, 38],
[118, 43, 41, 42, 144, 46, 18, 436, 18, 94],
[188, 98, 4, 8, 24, 9, 2, 8, 587, 72],
[ 75, 239, 8, 13, 12, 12, 10, 20, 43, 568]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 1s 7ms/step - loss: 2.7077 - accuracy: 0.2755 - val_loss: 2.3148 - val_accuracy: 0.3360 Epoch 2/200 92/92 [==============================] - 0s 5ms/step - loss: 2.1770 - accuracy: 0.3594 - val_loss: 2.0361 - val_accuracy: 0.3750 Epoch 3/200 92/92 [==============================] - 0s 5ms/step - loss: 1.9866 - accuracy: 0.3905 - val_loss: 1.8983 - val_accuracy: 0.4060 Epoch 4/200 92/92 [==============================] - 0s 5ms/step - loss: 1.8736 - accuracy: 0.4150 - val_loss: 1.8012 - val_accuracy: 0.4293 Epoch 5/200 92/92 [==============================] - 0s 5ms/step - loss: 1.8088 - accuracy: 0.4289 - val_loss: 1.7892 - val_accuracy: 0.4290 Epoch 6/200 92/92 [==============================] - 0s 5ms/step - loss: 1.7525 - accuracy: 0.4436 - val_loss: 1.7006 - val_accuracy: 0.4513 Epoch 7/200 92/92 [==============================] - 0s 5ms/step - loss: 1.7172 - accuracy: 0.4512 - val_loss: 1.6851 - val_accuracy: 0.4570 Epoch 8/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6739 - accuracy: 0.4630 - val_loss: 1.6636 - val_accuracy: 0.4647 Epoch 9/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6559 - accuracy: 0.4679 - val_loss: 1.6530 - val_accuracy: 0.4583 Epoch 10/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6333 - accuracy: 0.4750 - val_loss: 1.6764 - val_accuracy: 0.4567 Epoch 11/200 92/92 [==============================] - 0s 5ms/step - loss: 1.6118 - accuracy: 0.4817 - val_loss: 1.6240 - val_accuracy: 0.4723 Epoch 12/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5987 - accuracy: 0.4832 - val_loss: 1.6579 - val_accuracy: 0.4617 Epoch 13/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5928 - accuracy: 0.4853 - val_loss: 1.6160 - val_accuracy: 0.4823 Epoch 14/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5717 - accuracy: 0.4956 - val_loss: 1.5734 - val_accuracy: 0.5000 Epoch 15/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5641 - accuracy: 0.4940 - val_loss: 1.5605 - val_accuracy: 0.4910 Epoch 16/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5284 - accuracy: 0.5101 - val_loss: 1.5446 - val_accuracy: 0.5043 Epoch 17/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5356 - accuracy: 0.5043 - val_loss: 1.5549 - val_accuracy: 0.4957 Epoch 18/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5193 - accuracy: 0.5123 - val_loss: 1.5730 - val_accuracy: 0.4850 Epoch 19/200 92/92 [==============================] - 0s 5ms/step - loss: 1.5092 - accuracy: 0.5147 - val_loss: 1.5423 - val_accuracy: 0.5013
Model: "sequential_7" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= flatten_7 (Flatten) (None, 3072) 0 _________________________________________________________________ dense_18 (Dense) (None, 384) 1180032 _________________________________________________________________ dense_19 (Dense) (None, 192) 73920 _________________________________________________________________ dense_20 (Dense) (None, 96) 18528 _________________________________________________________________ dense_21 (Dense) (None, 10) 970 ================================================================= Total params: 1,273,450 Trainable params: 1,273,450 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 2ms/step - loss: 1.5492 - accuracy: 0.5023 test set accuracy: 50.23000240325928
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[529, 30, 88, 38, 21, 6, 24, 32, 186, 46],
[ 42, 615, 11, 41, 12, 5, 16, 24, 115, 119],
[ 58, 12, 386, 134, 112, 47, 117, 78, 35, 21],
[ 25, 23, 93, 422, 42, 101, 134, 57, 45, 58],
[ 50, 12, 165, 85, 388, 25, 122, 99, 41, 13],
[ 23, 12, 94, 312, 46, 261, 87, 84, 51, 30],
[ 5, 13, 78, 105, 113, 24, 592, 24, 24, 22],
[ 38, 15, 54, 96, 75, 38, 30, 571, 33, 50],
[ 81, 53, 15, 32, 13, 6, 13, 12, 717, 58],
[ 35, 185, 10, 48, 14, 11, 19, 43, 93, 542]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 1s 7ms/step - loss: 14.1347 - accuracy: 0.2492 - val_loss: 3.6442 - val_accuracy: 0.2600 Epoch 2/200 92/92 [==============================] - 0s 5ms/step - loss: 2.7269 - accuracy: 0.2624 - val_loss: 2.2832 - val_accuracy: 0.2737 Epoch 3/200 92/92 [==============================] - 0s 5ms/step - loss: 2.2308 - accuracy: 0.2528 - val_loss: 2.1515 - val_accuracy: 0.2850 Epoch 4/200 92/92 [==============================] - 0s 5ms/step - loss: 2.1600 - accuracy: 0.2582 - val_loss: 2.1831 - val_accuracy: 0.2313 Epoch 5/200 92/92 [==============================] - 0s 5ms/step - loss: 2.1374 - accuracy: 0.2523 - val_loss: 2.0936 - val_accuracy: 0.2683 Epoch 6/200 92/92 [==============================] - 0s 5ms/step - loss: 2.1204 - accuracy: 0.2519 - val_loss: 2.0971 - val_accuracy: 0.2703
Model: "sequential_8" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= flatten_8 (Flatten) (None, 3072) 0 _________________________________________________________________ dense_22 (Dense) (None, 384) 1180032 _________________________________________________________________ dense_23 (Dense) (None, 192) 73920 _________________________________________________________________ dense_24 (Dense) (None, 96) 18528 _________________________________________________________________ dense_25 (Dense) (None, 10) 970 ================================================================= Total params: 1,273,450 Trainable params: 1,273,450 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 2ms/step - loss: 2.1102 - accuracy: 0.2612 test set accuracy: 26.120001077651978
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[293, 80, 0, 6, 33, 149, 54, 49, 273, 63],
[ 32, 305, 1, 10, 33, 49, 194, 43, 79, 254],
[ 95, 58, 1, 3, 226, 161, 341, 59, 45, 11],
[ 37, 94, 0, 8, 223, 190, 344, 76, 15, 13],
[ 30, 53, 0, 3, 302, 115, 414, 52, 25, 6],
[ 49, 47, 0, 6, 257, 263, 291, 64, 18, 5],
[ 5, 68, 0, 4, 161, 64, 637, 55, 2, 4],
[ 46, 117, 1, 4, 246, 154, 276, 104, 22, 30],
[187, 104, 2, 7, 26, 86, 26, 32, 344, 186],
[ 38, 291, 0, 5, 14, 37, 133, 44, 83, 355]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 3s 31ms/step - loss: 2.0114 - accuracy: 0.3543 - val_loss: 1.5445 - val_accuracy: 0.4857 Epoch 2/200 92/92 [==============================] - 3s 30ms/step - loss: 1.4754 - accuracy: 0.5053 - val_loss: 1.3696 - val_accuracy: 0.5493 Epoch 3/200 92/92 [==============================] - 3s 30ms/step - loss: 1.3389 - accuracy: 0.5574 - val_loss: 1.2256 - val_accuracy: 0.6003 Epoch 4/200 92/92 [==============================] - 3s 30ms/step - loss: 1.2488 - accuracy: 0.5941 - val_loss: 1.1592 - val_accuracy: 0.6387 Epoch 5/200 92/92 [==============================] - 3s 29ms/step - loss: 1.1855 - accuracy: 0.6210 - val_loss: 1.0885 - val_accuracy: 0.6717 Epoch 6/200 92/92 [==============================] - 3s 29ms/step - loss: 1.1324 - accuracy: 0.6416 - val_loss: 1.0651 - val_accuracy: 0.6763 Epoch 7/200 92/92 [==============================] - 3s 29ms/step - loss: 1.0956 - accuracy: 0.6581 - val_loss: 1.0326 - val_accuracy: 0.6897 Epoch 8/200 92/92 [==============================] - 3s 29ms/step - loss: 1.0561 - accuracy: 0.6720 - val_loss: 1.0126 - val_accuracy: 0.6877 Epoch 9/200 92/92 [==============================] - 3s 29ms/step - loss: 1.0298 - accuracy: 0.6826 - val_loss: 0.9826 - val_accuracy: 0.7030 Epoch 10/200 92/92 [==============================] - 3s 30ms/step - loss: 0.9928 - accuracy: 0.6994 - val_loss: 0.9896 - val_accuracy: 0.7100 Epoch 11/200 92/92 [==============================] - 3s 29ms/step - loss: 0.9722 - accuracy: 0.7067 - val_loss: 0.9655 - val_accuracy: 0.7120 Epoch 12/200 92/92 [==============================] - 3s 30ms/step - loss: 0.9494 - accuracy: 0.7162 - val_loss: 0.9796 - val_accuracy: 0.7017 Epoch 13/200 92/92 [==============================] - 3s 30ms/step - loss: 0.9393 - accuracy: 0.7210 - val_loss: 0.9313 - val_accuracy: 0.7287 Epoch 14/200 92/92 [==============================] - 3s 30ms/step - loss: 0.9069 - accuracy: 0.7355 - val_loss: 0.9261 - val_accuracy: 0.7333 Epoch 15/200 92/92 [==============================] - 3s 29ms/step - loss: 0.8859 - accuracy: 0.7434 - val_loss: 0.9281 - val_accuracy: 0.7360 Epoch 16/200 92/92 [==============================] - 3s 29ms/step - loss: 0.8754 - accuracy: 0.7487 - val_loss: 0.9134 - val_accuracy: 0.7413 Epoch 17/200 92/92 [==============================] - 3s 30ms/step - loss: 0.8523 - accuracy: 0.7604 - val_loss: 0.8914 - val_accuracy: 0.7477 Epoch 18/200 92/92 [==============================] - 3s 30ms/step - loss: 0.8350 - accuracy: 0.7673 - val_loss: 0.9318 - val_accuracy: 0.7400 Epoch 19/200 92/92 [==============================] - 3s 29ms/step - loss: 0.8217 - accuracy: 0.7730 - val_loss: 0.8972 - val_accuracy: 0.7523 Epoch 20/200 92/92 [==============================] - 3s 30ms/step - loss: 0.8138 - accuracy: 0.7785 - val_loss: 0.8914 - val_accuracy: 0.7617 Epoch 21/200 92/92 [==============================] - 3s 30ms/step - loss: 0.7950 - accuracy: 0.7882 - val_loss: 0.8986 - val_accuracy: 0.7573 Epoch 22/200 92/92 [==============================] - 3s 29ms/step - loss: 0.7844 - accuracy: 0.7931 - val_loss: 0.9098 - val_accuracy: 0.7550 Epoch 23/200 92/92 [==============================] - 3s 29ms/step - loss: 0.7663 - accuracy: 0.8022 - val_loss: 0.8903 - val_accuracy: 0.7627 Epoch 24/200 92/92 [==============================] - 3s 29ms/step - loss: 0.7515 - accuracy: 0.8082 - val_loss: 0.9075 - val_accuracy: 0.7503 Epoch 25/200 92/92 [==============================] - 3s 29ms/step - loss: 0.7403 - accuracy: 0.8128 - val_loss: 0.9215 - val_accuracy: 0.7533 Epoch 26/200 92/92 [==============================] - 3s 30ms/step - loss: 0.7284 - accuracy: 0.8205 - val_loss: 0.9045 - val_accuracy: 0.7630 Epoch 27/200 92/92 [==============================] - 3s 30ms/step - loss: 0.7237 - accuracy: 0.8209 - val_loss: 0.9108 - val_accuracy: 0.7700 Epoch 28/200 92/92 [==============================] - 3s 30ms/step - loss: 0.7086 - accuracy: 0.8288 - val_loss: 0.9184 - val_accuracy: 0.7640 Epoch 29/200 92/92 [==============================] - 3s 29ms/step - loss: 0.6991 - accuracy: 0.8332 - val_loss: 0.9245 - val_accuracy: 0.7633 Epoch 30/200 92/92 [==============================] - 3s 29ms/step - loss: 0.6873 - accuracy: 0.8382 - val_loss: 0.9164 - val_accuracy: 0.7733 Epoch 31/200 92/92 [==============================] - 3s 29ms/step - loss: 0.6741 - accuracy: 0.8440 - val_loss: 0.9158 - val_accuracy: 0.7710 Epoch 32/200 92/92 [==============================] - 3s 29ms/step - loss: 0.6733 - accuracy: 0.8454 - val_loss: 0.9595 - val_accuracy: 0.7580 Epoch 33/200 92/92 [==============================] - 3s 29ms/step - loss: 0.6609 - accuracy: 0.8507 - val_loss: 0.9326 - val_accuracy: 0.7727
Model: "sequential_9" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_5 (Conv2D) (None, 30, 30, 128) 3584 _________________________________________________________________ max_pooling2d_5 (MaxPooling2 (None, 15, 15, 128) 0 _________________________________________________________________ dropout_5 (Dropout) (None, 15, 15, 128) 0 _________________________________________________________________ conv2d_6 (Conv2D) (None, 13, 13, 256) 295168 _________________________________________________________________ max_pooling2d_6 (MaxPooling2 (None, 6, 6, 256) 0 _________________________________________________________________ dropout_6 (Dropout) (None, 6, 6, 256) 0 _________________________________________________________________ flatten_9 (Flatten) (None, 9216) 0 _________________________________________________________________ dense_26 (Dense) (None, 384) 3539328 _________________________________________________________________ dense_27 (Dense) (None, 10) 3850 ================================================================= Total params: 3,841,930 Trainable params: 3,841,930 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 3ms/step - loss: 0.9610 - accuracy: 0.7601 test set accuracy: 76.010000705719
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[785, 19, 42, 14, 14, 6, 9, 9, 59, 43],
[ 9, 871, 10, 3, 2, 4, 12, 0, 18, 71],
[ 53, 4, 643, 63, 80, 50, 51, 25, 18, 13],
[ 16, 12, 59, 597, 43, 151, 48, 39, 14, 21],
[ 14, 2, 59, 68, 679, 31, 53, 75, 13, 6],
[ 10, 4, 48, 165, 29, 661, 18, 44, 10, 11],
[ 5, 2, 39, 64, 17, 20, 834, 6, 7, 6],
[ 11, 3, 23, 32, 22, 52, 5, 831, 3, 18],
[ 40, 33, 16, 12, 0, 3, 2, 5, 860, 29],
[ 17, 67, 11, 14, 3, 4, 4, 11, 29, 840]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 4s 38ms/step - loss: 2.1653 - accuracy: 0.2835 - val_loss: 1.7422 - val_accuracy: 0.3990 Epoch 2/200 92/92 [==============================] - 3s 36ms/step - loss: 1.5890 - accuracy: 0.4508 - val_loss: 1.4349 - val_accuracy: 0.5130 Epoch 3/200 92/92 [==============================] - 3s 35ms/step - loss: 1.3924 - accuracy: 0.5273 - val_loss: 1.2754 - val_accuracy: 0.5750 Epoch 4/200 92/92 [==============================] - 3s 36ms/step - loss: 1.2462 - accuracy: 0.5846 - val_loss: 1.1474 - val_accuracy: 0.6300 Epoch 5/200 92/92 [==============================] - 3s 36ms/step - loss: 1.1548 - accuracy: 0.6152 - val_loss: 1.0085 - val_accuracy: 0.6760 Epoch 6/200 92/92 [==============================] - 3s 35ms/step - loss: 1.0776 - accuracy: 0.6460 - val_loss: 1.0275 - val_accuracy: 0.6600 Epoch 7/200 92/92 [==============================] - 3s 35ms/step - loss: 1.0209 - accuracy: 0.6644 - val_loss: 0.9254 - val_accuracy: 0.6937 Epoch 8/200 92/92 [==============================] - 3s 35ms/step - loss: 0.9776 - accuracy: 0.6797 - val_loss: 0.9042 - val_accuracy: 0.7160 Epoch 9/200 92/92 [==============================] - 3s 35ms/step - loss: 0.9360 - accuracy: 0.6963 - val_loss: 0.8459 - val_accuracy: 0.7357 Epoch 10/200 92/92 [==============================] - 3s 35ms/step - loss: 0.9070 - accuracy: 0.7081 - val_loss: 0.8283 - val_accuracy: 0.7427 Epoch 11/200 92/92 [==============================] - 3s 36ms/step - loss: 0.8700 - accuracy: 0.7199 - val_loss: 0.7990 - val_accuracy: 0.7473 Epoch 12/200 92/92 [==============================] - 3s 36ms/step - loss: 0.8355 - accuracy: 0.7304 - val_loss: 0.7741 - val_accuracy: 0.7467 Epoch 13/200 92/92 [==============================] - 3s 36ms/step - loss: 0.8122 - accuracy: 0.7397 - val_loss: 0.7699 - val_accuracy: 0.7587 Epoch 14/200 92/92 [==============================] - 3s 36ms/step - loss: 0.7918 - accuracy: 0.7477 - val_loss: 0.7825 - val_accuracy: 0.7553 Epoch 15/200 92/92 [==============================] - 3s 35ms/step - loss: 0.7675 - accuracy: 0.7569 - val_loss: 0.7604 - val_accuracy: 0.7633 Epoch 16/200 92/92 [==============================] - 3s 36ms/step - loss: 0.7440 - accuracy: 0.7642 - val_loss: 0.7260 - val_accuracy: 0.7807 Epoch 17/200 92/92 [==============================] - 3s 36ms/step - loss: 0.7288 - accuracy: 0.7721 - val_loss: 0.7135 - val_accuracy: 0.7807 Epoch 18/200 92/92 [==============================] - 3s 36ms/step - loss: 0.7146 - accuracy: 0.7744 - val_loss: 0.7225 - val_accuracy: 0.7743 Epoch 19/200 92/92 [==============================] - 3s 35ms/step - loss: 0.6969 - accuracy: 0.7811 - val_loss: 0.6953 - val_accuracy: 0.7877 Epoch 20/200 92/92 [==============================] - 3s 35ms/step - loss: 0.6835 - accuracy: 0.7874 - val_loss: 0.6939 - val_accuracy: 0.7827 Epoch 21/200 92/92 [==============================] - 3s 35ms/step - loss: 0.6667 - accuracy: 0.7931 - val_loss: 0.6873 - val_accuracy: 0.7893 Epoch 22/200 92/92 [==============================] - 3s 36ms/step - loss: 0.6547 - accuracy: 0.7973 - val_loss: 0.6840 - val_accuracy: 0.7917 Epoch 23/200 92/92 [==============================] - 3s 36ms/step - loss: 0.6380 - accuracy: 0.8032 - val_loss: 0.7018 - val_accuracy: 0.7803 Epoch 24/200 92/92 [==============================] - 3s 36ms/step - loss: 0.6299 - accuracy: 0.8070 - val_loss: 0.6658 - val_accuracy: 0.7957 Epoch 25/200 92/92 [==============================] - 3s 36ms/step - loss: 0.6181 - accuracy: 0.8104 - val_loss: 0.6596 - val_accuracy: 0.8040 Epoch 26/200 92/92 [==============================] - 3s 35ms/step - loss: 0.6038 - accuracy: 0.8161 - val_loss: 0.6617 - val_accuracy: 0.7993 Epoch 27/200 92/92 [==============================] - 3s 36ms/step - loss: 0.5939 - accuracy: 0.8198 - val_loss: 0.6499 - val_accuracy: 0.8050 Epoch 28/200 92/92 [==============================] - 3s 36ms/step - loss: 0.5873 - accuracy: 0.8219 - val_loss: 0.6627 - val_accuracy: 0.8033 Epoch 29/200 92/92 [==============================] - 3s 36ms/step - loss: 0.5738 - accuracy: 0.8289 - val_loss: 0.6629 - val_accuracy: 0.7940 Epoch 30/200 92/92 [==============================] - 3s 36ms/step - loss: 0.5633 - accuracy: 0.8316 - val_loss: 0.6451 - val_accuracy: 0.8047
Model: "sequential_10" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_7 (Conv2D) (None, 30, 30, 128) 3584 _________________________________________________________________ max_pooling2d_7 (MaxPooling2 (None, 15, 15, 128) 0 _________________________________________________________________ dropout_7 (Dropout) (None, 15, 15, 128) 0 _________________________________________________________________ conv2d_8 (Conv2D) (None, 13, 13, 256) 295168 _________________________________________________________________ max_pooling2d_8 (MaxPooling2 (None, 6, 6, 256) 0 _________________________________________________________________ dropout_8 (Dropout) (None, 6, 6, 256) 0 _________________________________________________________________ conv2d_9 (Conv2D) (None, 4, 4, 512) 1180160 _________________________________________________________________ max_pooling2d_9 (MaxPooling2 (None, 2, 2, 512) 0 _________________________________________________________________ dropout_9 (Dropout) (None, 2, 2, 512) 0 _________________________________________________________________ flatten_10 (Flatten) (None, 2048) 0 _________________________________________________________________ dense_28 (Dense) (None, 384) 786816 _________________________________________________________________ dense_29 (Dense) (None, 10) 3850 ================================================================= Total params: 2,269,578 Trainable params: 2,269,578 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 3ms/step - loss: 0.6843 - accuracy: 0.7980 test set accuracy: 79.79999780654907
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[834, 23, 17, 18, 9, 1, 7, 11, 52, 28],
[ 8, 924, 2, 6, 0, 1, 4, 5, 15, 35],
[ 58, 5, 674, 60, 63, 37, 55, 29, 13, 6],
[ 12, 6, 51, 715, 40, 94, 38, 26, 8, 10],
[ 13, 2, 49, 61, 750, 23, 36, 54, 10, 2],
[ 12, 4, 32, 190, 34, 670, 12, 40, 2, 4],
[ 4, 4, 31, 55, 22, 13, 859, 6, 5, 1],
[ 8, 1, 24, 38, 24, 43, 4, 848, 1, 9],
[ 45, 21, 7, 15, 1, 3, 6, 6, 877, 19],
[ 15, 82, 6, 19, 1, 4, 6, 15, 23, 829]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 4s 37ms/step - loss: 6.8962 - accuracy: 0.2987 - val_loss: 1.7306 - val_accuracy: 0.3950 Epoch 2/200 92/92 [==============================] - 3s 36ms/step - loss: 1.6391 - accuracy: 0.4360 - val_loss: 1.5276 - val_accuracy: 0.4837 Epoch 3/200 92/92 [==============================] - 3s 35ms/step - loss: 1.4891 - accuracy: 0.5018 - val_loss: 1.4275 - val_accuracy: 0.5337 Epoch 4/200 92/92 [==============================] - 3s 36ms/step - loss: 1.4186 - accuracy: 0.5313 - val_loss: 1.3244 - val_accuracy: 0.5570 Epoch 5/200 92/92 [==============================] - 3s 36ms/step - loss: 1.3668 - accuracy: 0.5544 - val_loss: 1.2201 - val_accuracy: 0.6197 Epoch 6/200 92/92 [==============================] - 3s 36ms/step - loss: 1.3126 - accuracy: 0.5783 - val_loss: 1.2502 - val_accuracy: 0.5940 Epoch 7/200 92/92 [==============================] - 3s 36ms/step - loss: 1.2706 - accuracy: 0.5955 - val_loss: 1.2089 - val_accuracy: 0.6067 Epoch 8/200 92/92 [==============================] - 3s 36ms/step - loss: 1.2294 - accuracy: 0.6117 - val_loss: 1.1611 - val_accuracy: 0.6470 Epoch 9/200 92/92 [==============================] - 3s 36ms/step - loss: 1.1932 - accuracy: 0.6254 - val_loss: 1.1007 - val_accuracy: 0.6553 Epoch 10/200 92/92 [==============================] - 3s 36ms/step - loss: 1.1526 - accuracy: 0.6402 - val_loss: 1.0495 - val_accuracy: 0.6893 Epoch 11/200 92/92 [==============================] - 3s 36ms/step - loss: 1.1274 - accuracy: 0.6514 - val_loss: 1.0755 - val_accuracy: 0.6757 Epoch 12/200 92/92 [==============================] - 3s 35ms/step - loss: 1.1080 - accuracy: 0.6588 - val_loss: 1.0038 - val_accuracy: 0.7027 Epoch 13/200 92/92 [==============================] - 3s 36ms/step - loss: 1.0764 - accuracy: 0.6712 - val_loss: 0.9993 - val_accuracy: 0.7003 Epoch 14/200 92/92 [==============================] - 3s 36ms/step - loss: 1.0550 - accuracy: 0.6814 - val_loss: 1.0614 - val_accuracy: 0.6733 Epoch 15/200 92/92 [==============================] - 3s 35ms/step - loss: 1.0522 - accuracy: 0.6816 - val_loss: 0.9490 - val_accuracy: 0.7213 Epoch 16/200 92/92 [==============================] - 3s 35ms/step - loss: 1.0081 - accuracy: 0.6964 - val_loss: 0.9984 - val_accuracy: 0.6967 Epoch 17/200 92/92 [==============================] - 3s 36ms/step - loss: 0.9928 - accuracy: 0.7044 - val_loss: 0.9711 - val_accuracy: 0.7067 Epoch 18/200 92/92 [==============================] - 3s 36ms/step - loss: 0.9773 - accuracy: 0.7103 - val_loss: 0.9083 - val_accuracy: 0.7393 Epoch 19/200 92/92 [==============================] - 3s 36ms/step - loss: 0.9530 - accuracy: 0.7211 - val_loss: 0.8929 - val_accuracy: 0.7450 Epoch 20/200 92/92 [==============================] - 3s 35ms/step - loss: 0.9381 - accuracy: 0.7259 - val_loss: 0.8940 - val_accuracy: 0.7427 Epoch 21/200 92/92 [==============================] - 3s 36ms/step - loss: 0.9307 - accuracy: 0.7266 - val_loss: 0.8848 - val_accuracy: 0.7540 Epoch 22/200 92/92 [==============================] - 3s 35ms/step - loss: 0.9058 - accuracy: 0.7354 - val_loss: 0.8565 - val_accuracy: 0.7553 Epoch 23/200 92/92 [==============================] - 3s 35ms/step - loss: 0.8940 - accuracy: 0.7426 - val_loss: 0.8366 - val_accuracy: 0.7640 Epoch 24/200 92/92 [==============================] - 3s 35ms/step - loss: 0.8886 - accuracy: 0.7446 - val_loss: 0.8728 - val_accuracy: 0.7487 Epoch 25/200 92/92 [==============================] - 3s 36ms/step - loss: 0.8702 - accuracy: 0.7506 - val_loss: 0.8382 - val_accuracy: 0.7680 Epoch 26/200 92/92 [==============================] - 3s 35ms/step - loss: 0.8599 - accuracy: 0.7529 - val_loss: 0.8365 - val_accuracy: 0.7597 Epoch 27/200 92/92 [==============================] - 3s 36ms/step - loss: 0.8494 - accuracy: 0.7588 - val_loss: 0.8582 - val_accuracy: 0.7617 Epoch 28/200 92/92 [==============================] - 3s 35ms/step - loss: 0.8304 - accuracy: 0.7647 - val_loss: 0.8196 - val_accuracy: 0.7693 Epoch 29/200 92/92 [==============================] - 3s 36ms/step - loss: 0.8183 - accuracy: 0.7700 - val_loss: 0.8263 - val_accuracy: 0.7677 Epoch 30/200 92/92 [==============================] - 3s 35ms/step - loss: 0.8117 - accuracy: 0.7723 - val_loss: 0.7969 - val_accuracy: 0.7810 Epoch 31/200 92/92 [==============================] - 3s 35ms/step - loss: 0.8039 - accuracy: 0.7763 - val_loss: 0.7898 - val_accuracy: 0.7883 Epoch 32/200 92/92 [==============================] - 3s 35ms/step - loss: 0.7902 - accuracy: 0.7781 - val_loss: 0.7903 - val_accuracy: 0.7783 Epoch 33/200 92/92 [==============================] - 3s 36ms/step - loss: 0.7805 - accuracy: 0.7823 - val_loss: 0.7962 - val_accuracy: 0.7727 Epoch 34/200 92/92 [==============================] - 3s 35ms/step - loss: 0.7832 - accuracy: 0.7812 - val_loss: 0.7624 - val_accuracy: 0.7963 Epoch 35/200 92/92 [==============================] - 3s 36ms/step - loss: 0.7575 - accuracy: 0.7910 - val_loss: 0.7626 - val_accuracy: 0.7973 Epoch 36/200 92/92 [==============================] - 3s 38ms/step - loss: 0.7509 - accuracy: 0.7920 - val_loss: 0.7639 - val_accuracy: 0.7903 Epoch 37/200 92/92 [==============================] - 3s 35ms/step - loss: 0.7361 - accuracy: 0.7983 - val_loss: 0.7524 - val_accuracy: 0.7947 Epoch 38/200 92/92 [==============================] - 3s 35ms/step - loss: 0.7315 - accuracy: 0.8008 - val_loss: 0.7697 - val_accuracy: 0.7920
Model: "sequential_11" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_10 (Conv2D) (None, 30, 30, 128) 3584 _________________________________________________________________ max_pooling2d_10 (MaxPooling (None, 15, 15, 128) 0 _________________________________________________________________ dropout_10 (Dropout) (None, 15, 15, 128) 0 _________________________________________________________________ conv2d_11 (Conv2D) (None, 13, 13, 256) 295168 _________________________________________________________________ max_pooling2d_11 (MaxPooling (None, 6, 6, 256) 0 _________________________________________________________________ dropout_11 (Dropout) (None, 6, 6, 256) 0 _________________________________________________________________ conv2d_12 (Conv2D) (None, 4, 4, 512) 1180160 _________________________________________________________________ max_pooling2d_12 (MaxPooling (None, 2, 2, 512) 0 _________________________________________________________________ dropout_12 (Dropout) (None, 2, 2, 512) 0 _________________________________________________________________ flatten_11 (Flatten) (None, 2048) 0 _________________________________________________________________ dense_30 (Dense) (None, 384) 786816 _________________________________________________________________ dense_31 (Dense) (None, 10) 3850 ================================================================= Total params: 2,269,578 Trainable params: 2,269,578 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 3ms/step - loss: 0.8008 - accuracy: 0.7803 test set accuracy: 78.03000211715698
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[830, 7, 24, 23, 9, 3, 14, 12, 59, 19],
[ 18, 874, 2, 8, 0, 1, 19, 6, 29, 43],
[ 63, 4, 601, 91, 55, 58, 91, 28, 6, 3],
[ 10, 1, 38, 699, 32, 115, 68, 24, 8, 5],
[ 24, 1, 38, 78, 700, 29, 61, 60, 8, 1],
[ 11, 0, 23, 197, 28, 674, 15, 48, 2, 2],
[ 5, 1, 20, 63, 12, 13, 876, 6, 3, 1],
[ 13, 0, 12, 38, 36, 54, 6, 834, 1, 6],
[ 37, 13, 4, 16, 0, 5, 9, 5, 894, 17],
[ 24, 58, 9, 13, 3, 8, 17, 20, 27, 821]], dtype=int32)>
Epoch 1/200 735/735 [==============================] - 12s 15ms/step - loss: 1.6680 - accuracy: 0.4314 - val_loss: 1.2758 - val_accuracy: 0.5880 Epoch 2/200 735/735 [==============================] - 11s 15ms/step - loss: 1.2714 - accuracy: 0.5807 - val_loss: 1.1330 - val_accuracy: 0.6343 Epoch 3/200 735/735 [==============================] - 11s 15ms/step - loss: 1.1372 - accuracy: 0.6328 - val_loss: 1.0038 - val_accuracy: 0.6810 Epoch 4/200 735/735 [==============================] - 11s 15ms/step - loss: 1.0508 - accuracy: 0.6704 - val_loss: 0.9593 - val_accuracy: 0.7070 Epoch 5/200 735/735 [==============================] - 11s 15ms/step - loss: 0.9928 - accuracy: 0.6915 - val_loss: 0.9160 - val_accuracy: 0.7277 Epoch 6/200 735/735 [==============================] - 11s 15ms/step - loss: 0.9448 - accuracy: 0.7073 - val_loss: 0.8695 - val_accuracy: 0.7450 Epoch 7/200 735/735 [==============================] - 11s 15ms/step - loss: 0.8959 - accuracy: 0.7269 - val_loss: 0.8757 - val_accuracy: 0.7433 Epoch 8/200 735/735 [==============================] - 11s 15ms/step - loss: 0.8650 - accuracy: 0.7395 - val_loss: 0.7948 - val_accuracy: 0.7677 Epoch 9/200 735/735 [==============================] - 11s 15ms/step - loss: 0.8379 - accuracy: 0.7475 - val_loss: 0.8008 - val_accuracy: 0.7677 Epoch 10/200 735/735 [==============================] - 11s 15ms/step - loss: 0.8063 - accuracy: 0.7592 - val_loss: 0.7892 - val_accuracy: 0.7670 Epoch 11/200 735/735 [==============================] - 11s 15ms/step - loss: 0.7857 - accuracy: 0.7678 - val_loss: 0.7862 - val_accuracy: 0.7683 Epoch 12/200 735/735 [==============================] - 11s 15ms/step - loss: 0.7610 - accuracy: 0.7770 - val_loss: 0.7505 - val_accuracy: 0.7750 Epoch 13/200 735/735 [==============================] - 11s 15ms/step - loss: 0.7431 - accuracy: 0.7821 - val_loss: 0.7366 - val_accuracy: 0.7880 Epoch 14/200 735/735 [==============================] - 11s 15ms/step - loss: 0.7317 - accuracy: 0.7857 - val_loss: 0.7270 - val_accuracy: 0.7873 Epoch 15/200 735/735 [==============================] - 11s 15ms/step - loss: 0.7082 - accuracy: 0.7941 - val_loss: 0.7506 - val_accuracy: 0.7837 Epoch 16/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6990 - accuracy: 0.7971 - val_loss: 0.7391 - val_accuracy: 0.7897 Epoch 17/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6845 - accuracy: 0.8011 - val_loss: 0.7354 - val_accuracy: 0.7863 Epoch 18/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6665 - accuracy: 0.8076 - val_loss: 0.7036 - val_accuracy: 0.8040 Epoch 19/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6677 - accuracy: 0.8069 - val_loss: 0.7061 - val_accuracy: 0.7990 Epoch 20/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6449 - accuracy: 0.8144 - val_loss: 0.7050 - val_accuracy: 0.8017 Epoch 21/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6296 - accuracy: 0.8198 - val_loss: 0.6981 - val_accuracy: 0.8053 Epoch 22/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6174 - accuracy: 0.8249 - val_loss: 0.7008 - val_accuracy: 0.8053 Epoch 23/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6102 - accuracy: 0.8256 - val_loss: 0.6932 - val_accuracy: 0.8053 Epoch 24/200 735/735 [==============================] - 11s 15ms/step - loss: 0.6067 - accuracy: 0.8270 - val_loss: 0.7146 - val_accuracy: 0.7940 Epoch 25/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5978 - accuracy: 0.8293 - val_loss: 0.6727 - val_accuracy: 0.8063 Epoch 26/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5834 - accuracy: 0.8336 - val_loss: 0.7032 - val_accuracy: 0.7987 Epoch 27/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5777 - accuracy: 0.8339 - val_loss: 0.6653 - val_accuracy: 0.8153 Epoch 28/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5636 - accuracy: 0.8413 - val_loss: 0.6716 - val_accuracy: 0.8023 Epoch 29/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5640 - accuracy: 0.8403 - val_loss: 0.6821 - val_accuracy: 0.8087 Epoch 30/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5605 - accuracy: 0.8388 - val_loss: 0.6686 - val_accuracy: 0.8157 Epoch 31/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5471 - accuracy: 0.8447 - val_loss: 0.6637 - val_accuracy: 0.8163 Epoch 32/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5465 - accuracy: 0.8451 - val_loss: 0.6538 - val_accuracy: 0.8103 Epoch 33/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5334 - accuracy: 0.8490 - val_loss: 0.6667 - val_accuracy: 0.8030 Epoch 34/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5307 - accuracy: 0.8482 - val_loss: 0.7031 - val_accuracy: 0.7993 Epoch 35/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5199 - accuracy: 0.8522 - val_loss: 0.6531 - val_accuracy: 0.8160 Epoch 36/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5162 - accuracy: 0.8540 - val_loss: 0.6619 - val_accuracy: 0.8080 Epoch 37/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5117 - accuracy: 0.8559 - val_loss: 0.6616 - val_accuracy: 0.8103 Epoch 38/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4991 - accuracy: 0.8584 - val_loss: 0.6641 - val_accuracy: 0.8083 Epoch 39/200 735/735 [==============================] - 11s 15ms/step - loss: 0.5006 - accuracy: 0.8596 - val_loss: 0.6485 - val_accuracy: 0.8177 Epoch 40/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4935 - accuracy: 0.8626 - val_loss: 0.6577 - val_accuracy: 0.8133 Epoch 41/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4946 - accuracy: 0.8620 - val_loss: 0.6370 - val_accuracy: 0.8137 Epoch 42/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4829 - accuracy: 0.8641 - val_loss: 0.6500 - val_accuracy: 0.8097 Epoch 43/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4777 - accuracy: 0.8663 - val_loss: 0.6801 - val_accuracy: 0.8073 Epoch 44/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4818 - accuracy: 0.8638 - val_loss: 0.6777 - val_accuracy: 0.8050 Epoch 45/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4737 - accuracy: 0.8648 - val_loss: 0.6547 - val_accuracy: 0.8127 Epoch 46/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4630 - accuracy: 0.8686 - val_loss: 0.6294 - val_accuracy: 0.8210 Epoch 47/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4618 - accuracy: 0.8680 - val_loss: 0.6252 - val_accuracy: 0.8213 Epoch 48/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4537 - accuracy: 0.8725 - val_loss: 0.6488 - val_accuracy: 0.8137 Epoch 49/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4541 - accuracy: 0.8724 - val_loss: 0.6350 - val_accuracy: 0.8133 Epoch 50/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4488 - accuracy: 0.8736 - val_loss: 0.6326 - val_accuracy: 0.8233 Epoch 51/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4528 - accuracy: 0.8720 - val_loss: 0.6371 - val_accuracy: 0.8197 Epoch 52/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4444 - accuracy: 0.8734 - val_loss: 0.6311 - val_accuracy: 0.8203 Epoch 53/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4368 - accuracy: 0.8765 - val_loss: 0.6286 - val_accuracy: 0.8247 Epoch 54/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4380 - accuracy: 0.8760 - val_loss: 0.6409 - val_accuracy: 0.8177 Epoch 55/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4246 - accuracy: 0.8808 - val_loss: 0.6451 - val_accuracy: 0.8157 Epoch 56/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4214 - accuracy: 0.8815 - val_loss: 0.6735 - val_accuracy: 0.8083 Epoch 57/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4247 - accuracy: 0.8790 - val_loss: 0.6157 - val_accuracy: 0.8230 Epoch 58/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4153 - accuracy: 0.8816 - val_loss: 0.6321 - val_accuracy: 0.8247 Epoch 59/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4182 - accuracy: 0.8818 - val_loss: 0.6255 - val_accuracy: 0.8157 Epoch 60/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4161 - accuracy: 0.8834 - val_loss: 0.6361 - val_accuracy: 0.8157 Epoch 61/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4031 - accuracy: 0.8867 - val_loss: 0.6485 - val_accuracy: 0.8167 Epoch 62/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4073 - accuracy: 0.8846 - val_loss: 0.6351 - val_accuracy: 0.8180 Epoch 63/200 735/735 [==============================] - 11s 15ms/step - loss: 0.4052 - accuracy: 0.8848 - val_loss: 0.6439 - val_accuracy: 0.8093
Model: "sequential_18" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_37 (Conv2D) (None, 30, 30, 256) 7168 _________________________________________________________________ max_pooling2d_31 (MaxPooling (None, 15, 15, 256) 0 _________________________________________________________________ dropout_33 (Dropout) (None, 15, 15, 256) 0 _________________________________________________________________ conv2d_38 (Conv2D) (None, 13, 13, 512) 1180160 _________________________________________________________________ max_pooling2d_32 (MaxPooling (None, 6, 6, 512) 0 _________________________________________________________________ dropout_34 (Dropout) (None, 6, 6, 512) 0 _________________________________________________________________ conv2d_39 (Conv2D) (None, 4, 4, 640) 2949760 _________________________________________________________________ max_pooling2d_33 (MaxPooling (None, 2, 2, 640) 0 _________________________________________________________________ dropout_35 (Dropout) (None, 2, 2, 640) 0 _________________________________________________________________ flatten_18 (Flatten) (None, 2560) 0 _________________________________________________________________ dense_44 (Dense) (None, 480) 1229280 _________________________________________________________________ dense_45 (Dense) (None, 10) 4810 ================================================================= Total params: 5,371,178 Trainable params: 5,371,178 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 4ms/step - loss: 0.6893 - accuracy: 0.8014 test set accuracy: 80.14000058174133
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[745, 8, 29, 20, 32, 5, 7, 11, 95, 48],
[ 8, 870, 3, 2, 1, 7, 5, 3, 23, 78],
[ 58, 2, 668, 44, 99, 45, 39, 29, 8, 8],
[ 12, 3, 42, 637, 88, 129, 34, 26, 12, 17],
[ 8, 1, 27, 33, 857, 23, 14, 28, 8, 1],
[ 7, 3, 23, 112, 60, 740, 9, 31, 7, 8],
[ 3, 0, 29, 50, 51, 15, 839, 5, 6, 2],
[ 7, 1, 14, 27, 63, 39, 1, 829, 3, 16],
[ 14, 5, 4, 8, 4, 2, 1, 1, 928, 33],
[ 6, 27, 5, 13, 2, 10, 5, 3, 28, 901]], dtype=int32)>Model: "sequential_18" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_37 (Conv2D) (None, 30, 30, 256) 7168 _________________________________________________________________ max_pooling2d_31 (MaxPooling (None, 15, 15, 256) 0 _________________________________________________________________ dropout_33 (Dropout) (None, 15, 15, 256) 0 _________________________________________________________________ conv2d_38 (Conv2D) (None, 13, 13, 512) 1180160 _________________________________________________________________ max_pooling2d_32 (MaxPooling (None, 6, 6, 512) 0 _________________________________________________________________ dropout_34 (Dropout) (None, 6, 6, 512) 0 _________________________________________________________________ conv2d_39 (Conv2D) (None, 4, 4, 640) 2949760 _________________________________________________________________ max_pooling2d_33 (MaxPooling (None, 2, 2, 640) 0 _________________________________________________________________ dropout_35 (Dropout) (None, 2, 2, 640) 0 _________________________________________________________________ flatten_18 (Flatten) (None, 2560) 0 _________________________________________________________________ dense_44 (Dense) (None, 480) 1229280 _________________________________________________________________ dense_45 (Dense) (None, 10) 4810 ================================================================= Total params: 5,371,178 Trainable params: 5,371,178 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 1s 4ms/step - loss: 0.6893 - accuracy: 0.8014 test set accuracy: 80.14000058174133
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[745, 8, 29, 20, 32, 5, 7, 11, 95, 48],
[ 8, 870, 3, 2, 1, 7, 5, 3, 23, 78],
[ 58, 2, 668, 44, 99, 45, 39, 29, 8, 8],
[ 12, 3, 42, 637, 88, 129, 34, 26, 12, 17],
[ 8, 1, 27, 33, 857, 23, 14, 28, 8, 1],
[ 7, 3, 23, 112, 60, 740, 9, 31, 7, 8],
[ 3, 0, 29, 50, 51, 15, 839, 5, 6, 2],
[ 7, 1, 14, 27, 63, 39, 1, 829, 3, 16],
[ 14, 5, 4, 8, 4, 2, 1, 1, 928, 33],
[ 6, 27, 5, 13, 2, 10, 5, 3, 28, 901]], dtype=int32)>8
['conv2d_40', 'max_pooling2d_34', 'dropout_36', 'conv2d_41', 'max_pooling2d_35', 'dropout_37', 'conv2d_42', 'max_pooling2d_36', 'dropout_38', 'flatten_19', 'dense_46', 'dense_47']
Epoch 1/200 92/92 [==============================] - 46s 447ms/step - loss: 3.3636 - accuracy: 0.2189 - val_loss: 2.2945 - val_accuracy: 0.3137 Epoch 2/200 92/92 [==============================] - 35s 381ms/step - loss: 1.9718 - accuracy: 0.3782 - val_loss: 1.7066 - val_accuracy: 0.4337 Epoch 3/200 92/92 [==============================] - 35s 382ms/step - loss: 1.6649 - accuracy: 0.4390 - val_loss: 1.5279 - val_accuracy: 0.4803 Epoch 4/200 92/92 [==============================] - 35s 382ms/step - loss: 1.5673 - accuracy: 0.4633 - val_loss: 1.4713 - val_accuracy: 0.4990 Epoch 5/200 92/92 [==============================] - 35s 382ms/step - loss: 1.4997 - accuracy: 0.4872 - val_loss: 1.4051 - val_accuracy: 0.5207 Epoch 6/200 92/92 [==============================] - 35s 382ms/step - loss: 1.4526 - accuracy: 0.5078 - val_loss: 1.4108 - val_accuracy: 0.5203 Epoch 7/200 92/92 [==============================] - 35s 382ms/step - loss: 1.4141 - accuracy: 0.5226 - val_loss: 1.4223 - val_accuracy: 0.5227 Epoch 8/200 92/92 [==============================] - 35s 382ms/step - loss: 1.3930 - accuracy: 0.5317 - val_loss: 1.3775 - val_accuracy: 0.5403 Epoch 9/200 92/92 [==============================] - 35s 383ms/step - loss: 1.3670 - accuracy: 0.5442 - val_loss: 1.3765 - val_accuracy: 0.5357 Epoch 10/200 92/92 [==============================] - 35s 383ms/step - loss: 1.3415 - accuracy: 0.5543 - val_loss: 1.3778 - val_accuracy: 0.5377 Epoch 11/200 92/92 [==============================] - 35s 382ms/step - loss: 1.3314 - accuracy: 0.5613 - val_loss: 1.3469 - val_accuracy: 0.5513 Epoch 12/200 92/92 [==============================] - 35s 383ms/step - loss: 1.3079 - accuracy: 0.5730 - val_loss: 1.3687 - val_accuracy: 0.5433 Epoch 13/200 92/92 [==============================] - 35s 382ms/step - loss: 1.2977 - accuracy: 0.5764 - val_loss: 1.3320 - val_accuracy: 0.5633 Epoch 14/200 92/92 [==============================] - 35s 383ms/step - loss: 1.2759 - accuracy: 0.5847 - val_loss: 1.3244 - val_accuracy: 0.5653 Epoch 15/200 92/92 [==============================] - 35s 382ms/step - loss: 1.2644 - accuracy: 0.5944 - val_loss: 1.3121 - val_accuracy: 0.5697 Epoch 16/200 92/92 [==============================] - 35s 383ms/step - loss: 1.2469 - accuracy: 0.6038 - val_loss: 1.3272 - val_accuracy: 0.5683 Epoch 17/200 92/92 [==============================] - 35s 383ms/step - loss: 1.2488 - accuracy: 0.5994 - val_loss: 1.3092 - val_accuracy: 0.5750 Epoch 18/200 92/92 [==============================] - 35s 383ms/step - loss: 1.2187 - accuracy: 0.6116 - val_loss: 1.3583 - val_accuracy: 0.5607 Epoch 19/200 92/92 [==============================] - 35s 383ms/step - loss: 1.2155 - accuracy: 0.6164 - val_loss: 1.3062 - val_accuracy: 0.5763 Epoch 20/200 92/92 [==============================] - 35s 383ms/step - loss: 1.1972 - accuracy: 0.6255 - val_loss: 1.3147 - val_accuracy: 0.5793 Epoch 21/200 92/92 [==============================] - 35s 383ms/step - loss: 1.1921 - accuracy: 0.6280 - val_loss: 1.4167 - val_accuracy: 0.5660 Epoch 22/200 92/92 [==============================] - 35s 383ms/step - loss: 1.1762 - accuracy: 0.6337 - val_loss: 1.3526 - val_accuracy: 0.5723 Epoch 23/200 92/92 [==============================] - 35s 382ms/step - loss: 1.1649 - accuracy: 0.6396 - val_loss: 1.3091 - val_accuracy: 0.5867 Epoch 24/200 92/92 [==============================] - 35s 383ms/step - loss: 1.1677 - accuracy: 0.6416 - val_loss: 1.3813 - val_accuracy: 0.5753 Epoch 25/200 92/92 [==============================] - 35s 383ms/step - loss: 1.1499 - accuracy: 0.6479 - val_loss: 1.3226 - val_accuracy: 0.5960 Epoch 26/200 92/92 [==============================] - 35s 382ms/step - loss: 1.1347 - accuracy: 0.6544 - val_loss: 1.3575 - val_accuracy: 0.5917 Epoch 27/200 92/92 [==============================] - 35s 382ms/step - loss: 1.1373 - accuracy: 0.6552 - val_loss: 1.3540 - val_accuracy: 0.5903 Epoch 28/200 92/92 [==============================] - 35s 383ms/step - loss: 1.1242 - accuracy: 0.6622 - val_loss: 1.3707 - val_accuracy: 0.5780 Epoch 29/200 92/92 [==============================] - 35s 383ms/step - loss: 1.1189 - accuracy: 0.6639 - val_loss: 1.3683 - val_accuracy: 0.5873 Epoch 30/200 92/92 [==============================] - 35s 382ms/step - loss: 1.1060 - accuracy: 0.6707 - val_loss: 1.3549 - val_accuracy: 0.5970 Epoch 31/200 92/92 [==============================] - 35s 383ms/step - loss: 1.0888 - accuracy: 0.6794 - val_loss: 1.3490 - val_accuracy: 0.6033 Epoch 32/200 92/92 [==============================] - 35s 383ms/step - loss: 1.0880 - accuracy: 0.6806 - val_loss: 1.3795 - val_accuracy: 0.5900 Epoch 33/200 92/92 [==============================] - 35s 383ms/step - loss: 1.0802 - accuracy: 0.6827 - val_loss: 1.4350 - val_accuracy: 0.5797 Epoch 34/200 92/92 [==============================] - 35s 383ms/step - loss: 1.0751 - accuracy: 0.6875 - val_loss: 1.3973 - val_accuracy: 0.5810 Epoch 35/200 92/92 [==============================] - 35s 382ms/step - loss: 1.0636 - accuracy: 0.6946 - val_loss: 1.4155 - val_accuracy: 0.5900 Epoch 36/200 92/92 [==============================] - 35s 382ms/step - loss: 1.0575 - accuracy: 0.6967 - val_loss: 1.4249 - val_accuracy: 0.5840 Epoch 37/200 92/92 [==============================] - 36s 386ms/step - loss: 1.0442 - accuracy: 0.6995 - val_loss: 1.4189 - val_accuracy: 0.5897 Epoch 38/200 92/92 [==============================] - 35s 383ms/step - loss: 1.0311 - accuracy: 0.7052 - val_loss: 1.4203 - val_accuracy: 0.5903 Epoch 39/200 92/92 [==============================] - 35s 382ms/step - loss: 1.0324 - accuracy: 0.7074 - val_loss: 1.4851 - val_accuracy: 0.5813 Epoch 40/200 92/92 [==============================] - 35s 383ms/step - loss: 1.0393 - accuracy: 0.7075 - val_loss: 1.4486 - val_accuracy: 0.5873 Epoch 41/200 92/92 [==============================] - 35s 383ms/step - loss: 1.0291 - accuracy: 0.7096 - val_loss: 1.4829 - val_accuracy: 0.5827
Model: "sequential_14" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_22 (Conv2D) (None, 31, 31, 128) 1664 _________________________________________________________________ max_pooling2d_19 (MaxPooling (None, 31, 31, 128) 0 _________________________________________________________________ dropout_20 (Dropout) (None, 31, 31, 128) 0 _________________________________________________________________ conv2d_23 (Conv2D) (None, 30, 30, 256) 131328 _________________________________________________________________ max_pooling2d_20 (MaxPooling (None, 30, 30, 256) 0 _________________________________________________________________ dropout_21 (Dropout) (None, 30, 30, 256) 0 _________________________________________________________________ conv2d_24 (Conv2D) (None, 29, 29, 512) 524800 _________________________________________________________________ max_pooling2d_21 (MaxPooling (None, 29, 29, 512) 0 _________________________________________________________________ dropout_22 (Dropout) (None, 29, 29, 512) 0 _________________________________________________________________ flatten_14 (Flatten) (None, 430592) 0 _________________________________________________________________ dense_36 (Dense) (None, 384) 165347712 _________________________________________________________________ dense_37 (Dense) (None, 10) 3850 ================================================================= Total params: 166,009,354 Trainable params: 166,009,354 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 5s 14ms/step - loss: 1.5885 - accuracy: 0.5565 test set accuracy: 55.650001764297485
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[629, 22, 113, 31, 30, 9, 23, 13, 96, 34],
[ 37, 706, 6, 41, 10, 18, 14, 16, 59, 93],
[ 46, 6, 471, 103, 181, 42, 86, 36, 23, 6],
[ 16, 8, 88, 410, 150, 95, 155, 38, 23, 17],
[ 19, 0, 98, 97, 570, 34, 109, 60, 9, 4],
[ 11, 7, 98, 285, 114, 281, 115, 72, 10, 7],
[ 3, 8, 39, 98, 103, 15, 708, 10, 10, 6],
[ 14, 5, 51, 99, 129, 68, 34, 574, 10, 16],
[128, 39, 51, 58, 18, 12, 16, 12, 645, 21],
[ 49, 90, 16, 94, 29, 23, 40, 32, 56, 571]], dtype=int32)>
Epoch 1/200 92/92 [==============================] - 13s 119ms/step - loss: 1.9704 - accuracy: 0.2837 - val_loss: 1.5952 - val_accuracy: 0.4187 Epoch 2/200 92/92 [==============================] - 8s 88ms/step - loss: 1.4909 - accuracy: 0.4687 - val_loss: 1.3238 - val_accuracy: 0.5360 Epoch 3/200 92/92 [==============================] - 8s 88ms/step - loss: 1.3041 - accuracy: 0.5474 - val_loss: 1.1958 - val_accuracy: 0.5920 Epoch 4/200 92/92 [==============================] - 8s 88ms/step - loss: 1.1770 - accuracy: 0.5968 - val_loss: 1.0476 - val_accuracy: 0.6567 Epoch 5/200 92/92 [==============================] - 8s 88ms/step - loss: 1.0772 - accuracy: 0.6311 - val_loss: 0.9850 - val_accuracy: 0.6773 Epoch 6/200 92/92 [==============================] - 8s 88ms/step - loss: 1.0108 - accuracy: 0.6578 - val_loss: 0.9322 - val_accuracy: 0.6920 Epoch 7/200 92/92 [==============================] - 8s 88ms/step - loss: 0.9583 - accuracy: 0.6776 - val_loss: 0.8516 - val_accuracy: 0.7267 Epoch 8/200 92/92 [==============================] - 8s 88ms/step - loss: 0.8963 - accuracy: 0.6998 - val_loss: 0.8104 - val_accuracy: 0.7393 Epoch 9/200 92/92 [==============================] - 8s 88ms/step - loss: 0.8570 - accuracy: 0.7169 - val_loss: 0.8398 - val_accuracy: 0.7340 Epoch 10/200 92/92 [==============================] - 8s 88ms/step - loss: 0.8273 - accuracy: 0.7264 - val_loss: 0.7797 - val_accuracy: 0.7500 Epoch 11/200 92/92 [==============================] - 8s 88ms/step - loss: 0.7802 - accuracy: 0.7416 - val_loss: 0.7468 - val_accuracy: 0.7690 Epoch 12/200 92/92 [==============================] - 8s 88ms/step - loss: 0.7483 - accuracy: 0.7539 - val_loss: 0.7221 - val_accuracy: 0.7720 Epoch 13/200 92/92 [==============================] - 8s 88ms/step - loss: 0.7185 - accuracy: 0.7670 - val_loss: 0.7534 - val_accuracy: 0.7563 Epoch 14/200 92/92 [==============================] - 8s 88ms/step - loss: 0.6979 - accuracy: 0.7731 - val_loss: 0.7291 - val_accuracy: 0.7720 Epoch 15/200 92/92 [==============================] - 8s 88ms/step - loss: 0.6816 - accuracy: 0.7788 - val_loss: 0.7192 - val_accuracy: 0.7733 Epoch 16/200 92/92 [==============================] - 8s 88ms/step - loss: 0.6573 - accuracy: 0.7889 - val_loss: 0.6892 - val_accuracy: 0.7803 Epoch 17/200 92/92 [==============================] - 8s 88ms/step - loss: 0.6305 - accuracy: 0.7979 - val_loss: 0.6888 - val_accuracy: 0.7907 Epoch 18/200 92/92 [==============================] - 8s 88ms/step - loss: 0.6132 - accuracy: 0.8036 - val_loss: 0.7068 - val_accuracy: 0.7790 Epoch 19/200 92/92 [==============================] - 8s 88ms/step - loss: 0.6047 - accuracy: 0.8076 - val_loss: 0.6800 - val_accuracy: 0.7887 Epoch 20/200 92/92 [==============================] - 8s 88ms/step - loss: 0.5821 - accuracy: 0.8160 - val_loss: 0.6849 - val_accuracy: 0.7803 Epoch 21/200 92/92 [==============================] - 8s 88ms/step - loss: 0.5570 - accuracy: 0.8251 - val_loss: 0.6717 - val_accuracy: 0.7897 Epoch 22/200 92/92 [==============================] - 8s 88ms/step - loss: 0.5409 - accuracy: 0.8321 - val_loss: 0.6633 - val_accuracy: 0.8007 Epoch 23/200 92/92 [==============================] - 8s 88ms/step - loss: 0.5350 - accuracy: 0.8335 - val_loss: 0.6592 - val_accuracy: 0.7960 Epoch 24/200 92/92 [==============================] - 8s 88ms/step - loss: 0.5083 - accuracy: 0.8423 - val_loss: 0.6586 - val_accuracy: 0.8010 Epoch 25/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4993 - accuracy: 0.8474 - val_loss: 0.6549 - val_accuracy: 0.7997 Epoch 26/200 92/92 [==============================] - 8s 88ms/step - loss: 0.5041 - accuracy: 0.8465 - val_loss: 0.6829 - val_accuracy: 0.7930 Epoch 27/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4815 - accuracy: 0.8538 - val_loss: 0.6603 - val_accuracy: 0.8020 Epoch 28/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4683 - accuracy: 0.8579 - val_loss: 0.6643 - val_accuracy: 0.8017 Epoch 29/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4591 - accuracy: 0.8633 - val_loss: 0.6608 - val_accuracy: 0.7927 Epoch 30/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4462 - accuracy: 0.8676 - val_loss: 0.6713 - val_accuracy: 0.8043 Epoch 31/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4416 - accuracy: 0.8695 - val_loss: 0.6637 - val_accuracy: 0.8070 Epoch 32/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4266 - accuracy: 0.8753 - val_loss: 0.6774 - val_accuracy: 0.8040 Epoch 33/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4191 - accuracy: 0.8774 - val_loss: 0.6679 - val_accuracy: 0.8017 Epoch 34/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4155 - accuracy: 0.8814 - val_loss: 0.6854 - val_accuracy: 0.8003 Epoch 35/200 92/92 [==============================] - 8s 88ms/step - loss: 0.4000 - accuracy: 0.8866 - val_loss: 0.6808 - val_accuracy: 0.8020 Epoch 36/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3986 - accuracy: 0.8872 - val_loss: 0.6827 - val_accuracy: 0.8047 Epoch 37/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3860 - accuracy: 0.8920 - val_loss: 0.6774 - val_accuracy: 0.8027 Epoch 38/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3793 - accuracy: 0.8954 - val_loss: 0.6856 - val_accuracy: 0.8050 Epoch 39/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3710 - accuracy: 0.8961 - val_loss: 0.6902 - val_accuracy: 0.8067 Epoch 40/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3707 - accuracy: 0.9002 - val_loss: 0.6739 - val_accuracy: 0.8087 Epoch 41/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3657 - accuracy: 0.9014 - val_loss: 0.7075 - val_accuracy: 0.8090 Epoch 42/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3659 - accuracy: 0.9013 - val_loss: 0.7132 - val_accuracy: 0.8080 Epoch 43/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3440 - accuracy: 0.9102 - val_loss: 0.7143 - val_accuracy: 0.8073 Epoch 44/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3433 - accuracy: 0.9092 - val_loss: 0.6964 - val_accuracy: 0.8097 Epoch 45/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3361 - accuracy: 0.9133 - val_loss: 0.7172 - val_accuracy: 0.8003 Epoch 46/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3419 - accuracy: 0.9118 - val_loss: 0.7165 - val_accuracy: 0.8107 Epoch 47/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3347 - accuracy: 0.9145 - val_loss: 0.7256 - val_accuracy: 0.8070 Epoch 48/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3311 - accuracy: 0.9158 - val_loss: 0.7293 - val_accuracy: 0.8110 Epoch 49/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3348 - accuracy: 0.9158 - val_loss: 0.7216 - val_accuracy: 0.8090 Epoch 50/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3243 - accuracy: 0.9187 - val_loss: 0.7128 - val_accuracy: 0.8117 Epoch 51/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3192 - accuracy: 0.9229 - val_loss: 0.7343 - val_accuracy: 0.8057 Epoch 52/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3234 - accuracy: 0.9205 - val_loss: 0.7181 - val_accuracy: 0.8143 Epoch 53/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3147 - accuracy: 0.9242 - val_loss: 0.7392 - val_accuracy: 0.8090 Epoch 54/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3101 - accuracy: 0.9257 - val_loss: 0.7448 - val_accuracy: 0.8113 Epoch 55/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3068 - accuracy: 0.9264 - val_loss: 0.7618 - val_accuracy: 0.8000 Epoch 56/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3054 - accuracy: 0.9281 - val_loss: 0.7329 - val_accuracy: 0.8070 Epoch 57/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3074 - accuracy: 0.9276 - val_loss: 0.7236 - val_accuracy: 0.8170 Epoch 58/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3080 - accuracy: 0.9265 - val_loss: 0.7443 - val_accuracy: 0.8100 Epoch 59/200 92/92 [==============================] - 8s 88ms/step - loss: 0.2957 - accuracy: 0.9316 - val_loss: 0.7611 - val_accuracy: 0.8067 Epoch 60/200 92/92 [==============================] - 8s 88ms/step - loss: 0.3038 - accuracy: 0.9294 - val_loss: 0.7704 - val_accuracy: 0.8043 Epoch 61/200 92/92 [==============================] - 8s 88ms/step - loss: 0.2909 - accuracy: 0.9337 - val_loss: 0.7742 - val_accuracy: 0.8107 Epoch 62/200 92/92 [==============================] - 8s 88ms/step - loss: 0.2954 - accuracy: 0.9334 - val_loss: 0.7745 - val_accuracy: 0.8033 Epoch 63/200 92/92 [==============================] - 8s 87ms/step - loss: 0.2977 - accuracy: 0.9317 - val_loss: 0.7697 - val_accuracy: 0.7990 Epoch 64/200 92/92 [==============================] - 8s 88ms/step - loss: 0.2869 - accuracy: 0.9363 - val_loss: 0.7729 - val_accuracy: 0.8080 Epoch 65/200 92/92 [==============================] - 8s 88ms/step - loss: 0.2955 - accuracy: 0.9336 - val_loss: 0.7796 - val_accuracy: 0.8060 Epoch 66/200 92/92 [==============================] - 8s 88ms/step - loss: 0.2877 - accuracy: 0.9371 - val_loss: 0.7631 - val_accuracy: 0.8027 Epoch 67/200 92/92 [==============================] - 8s 88ms/step - loss: 0.2796 - accuracy: 0.9399 - val_loss: 0.7857 - val_accuracy: 0.8063
Model: "sequential_15" _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= conv2d_25 (Conv2D) (None, 30, 30, 256) 7168 _________________________________________________________________ max_pooling2d_22 (MaxPooling (None, 15, 15, 256) 0 _________________________________________________________________ dropout_23 (Dropout) (None, 15, 15, 256) 0 _________________________________________________________________ conv2d_26 (Conv2D) (None, 13, 13, 512) 1180160 _________________________________________________________________ max_pooling2d_23 (MaxPooling (None, 6, 6, 512) 0 _________________________________________________________________ dropout_24 (Dropout) (None, 6, 6, 512) 0 _________________________________________________________________ conv2d_27 (Conv2D) (None, 4, 4, 1024) 4719616 _________________________________________________________________ max_pooling2d_24 (MaxPooling (None, 2, 2, 1024) 0 _________________________________________________________________ dropout_25 (Dropout) (None, 2, 2, 1024) 0 _________________________________________________________________ flatten_15 (Flatten) (None, 4096) 0 _________________________________________________________________ dense_38 (Dense) (None, 384) 1573248 _________________________________________________________________ dense_39 (Dense) (None, 10) 3850 ================================================================= Total params: 7,484,042 Trainable params: 7,484,042 Non-trainable params: 0 _________________________________________________________________
313/313 [==============================] - 2s 5ms/step - loss: 0.8150 - accuracy: 0.7968 test set accuracy: 79.68000173568726
shape of preds: (10000, 10)
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])
<tf.Tensor: shape=(10, 10), dtype=int32, numpy=
array([[835, 21, 27, 7, 11, 3, 6, 9, 60, 21],
[ 9, 924, 1, 0, 3, 2, 4, 0, 21, 36],
[ 55, 5, 682, 58, 63, 46, 38, 25, 24, 4],
[ 22, 13, 54, 567, 61, 155, 44, 44, 24, 16],
[ 11, 9, 50, 35, 791, 24, 24, 44, 10, 2],
[ 11, 4, 33, 127, 39, 706, 14, 47, 10, 9],
[ 9, 5, 34, 26, 28, 17, 861, 5, 9, 6],
[ 17, 4, 15, 16, 40, 33, 1, 854, 9, 11],
[ 34, 27, 9, 5, 3, 0, 3, 2, 907, 10],
[ 20, 88, 4, 4, 3, 2, 3, 3, 32, 841]], dtype=int32)>